Major fire at Sandoz chemicals warehouse 956 in Schweizerhalle
During the night of Friday 31 October to Saturday 1 November 1986, a fire broke out at Sandoz warehouse 956 on the industrial site at Schweizerhalle near Basel (Switzerland). Investigations carried out by the scientific branch of the Zurich city police later showed that the fire had probably originated during the packaging of the inorganic pigment Prussian blue (iron(III) hexacyanoferrate(II), Fe
[FeII(CN)6]3) (Hurni 1988). In accordance with the then-current state of packaging technology, the pigment was covered with a plastic sheet and shrink-wrapped using a blowtorch. The warehousemen did not realize that the naked flame had ignited the packaged material. Some hours later, the glowing process developed into the fateful fire (Fig. 1a). In fire tests conducted after the event, Prussian blue was found to be “readily ignitable” and to burn with a “flameless, smokeless, slowly progressing glowing”. These properties were, however, also described in the relevant substance documentation. The fire was discovered shortly after midnight, and within a matter of minutes the company and public fire services were attempting to extinguish the blaze and protect the surrounding area. Despite the considerable risks, the fire fighters, around 160 in all, were selfless in their efforts, as noted by the then Basel correspondent of the Neue Zürcher Zeitung in his report: “The horrific nature of the blaze was quite staggering. Again and again, exploding drums and eruptions of fire sent sheets of flame high up into the night sky, and the firemen in protective masks had a hard time of it fighting this unleashed element, especially early on.” Figure 1a and b give an impression of the situation that night.
In the early hours of the morning of 1 November 1986, at 3.40 am, the fire had been extinguished. It would have been possible to allow the warehouse to burn down completely, while maintaining “cold walls” to protect adjacent buildings. However, as one of the substances stored in a neighboring warehouse was phosgene and this highly potent poisonous gas had to be prevented from escaping at all costs, the priority for the fire services was to put out the warehouse fire as quickly as possible (Preiswerk 2005). A very high additional risk was posed by metallic sodium stored nearby (Hurni 1988), given the large amounts of water used to fight the fire (the high temperatures prevented the use of foam). By the time the fire was extinguished, the warehouse had collapsed and the scene was one of chaotic devastation (Fig. 1c, d).
One highly unfavorable circumstance was the fact that warehouse 956, measuring 90 by 50 m, had originally been constructed to store machinery and was only later converted for storage of chemicals. This explains, among other things, why the building lacked adequate smoke detection and sprinkler systems and only contained one dividing wall. Stock record-keeping procedures also turned out to have serious deficiencies, and as a result the stock lists published by Sandoz after the fire were amended several times. Altogether, 1,250 t of agrochemicals and intermediates were stored in warehouse 956. A summary of the products and the most important active ingredients is given in Table 1. The main substances were phosphoric acid ester (organophosphorus) insecticides, e.g., disulfoton, thiometon, and etrimphos (Capel et al. 1988; DK-Rhein 1986). Highly toxic mercury compounds were present as fungicides in seed dressings and were marked with the fluorescent red dye rhodamine B, which in turn was responsible for the reddening of the fire-fighting water and subsequently of the Rhine (Fig. 1e).
Emissions and damage to the environment
The bulk of the chemicals stored in warehouse 956 were consumed by the fire, with the prevailing temperatures also causing the steel girders to melt (Fig. 1c). It was not possible to determine what combustion products were generated in the process. On the basis of the malodorous smoke, it was assumed that sulfurous compounds (mercaptans) were significant components.
The population of the Basel region was rudely awakened to the disaster in the early hours of the morning. As a foul-smelling gas cloud from the blaze was spreading across the conurbation, the authorities activated an emergency alert. In Kleinbasel and the communes of the Canton Basel-Landschaft the sirens wailed, and in Grossbasel police loudspeaker vans patrolled the streets, instructing residents to keep their doors and windows closed and to stay indoors. The situation was alarming not least because the information broadcast via local radio to residents unable to leave their homes was incomplete and contradictory, and because for some hours they did not know how dangerous the malodorous air pollution actually was. The all clear was not given until 7 am. As the gas cloud was known—so it was claimed—not to contain any toxic substances, no lasting damage to health was to be expected.
In the early morning of 1 November 1986 in the Basel region, there was prolonged uncertainty as to whether the consequences of the Sandoz fire would be as catastrophic as those of the accidents that had occurred at Bhopal, Seveso or—in April of the same year—Chernobyl. It is not clear, to this author at least, how the risks posed by the air pollution could have been assessed with sufficient certainty.
It was only reported later that the substances stored in warehouse 956 included 2.3 t of tetradifon, an agent used to kill mites (acaricide). Since tetradifon (4-chlorophenyl 2,4,5-trichlorophenyl sulfone) via 1,2,4-trichlorophenol is readily identifiable as a precursor for the formation of dioxins, especially the highly toxic 2,3,7,8-tetrachlorodibenzo-p-dioxin, soil samples and the activated carbon filters of fire fighters’ protective masks were subsequently analyzed for polychlorinated dibenzodioxins and dibenzofurans in an extended measurement program. Fortunately, no elevated concentrations were detected.
Given the rapid dynamics of air pollution, it was not possible to carry out detailed chemical analyses of air samples, such as were later performed on samples of Rhine water and sediments, groundwater, and drinking water as well as at the site of the fire and in the underlying soil. The risk of air pollution from an industrial fire remains relatively high, as was clearly demonstrated, for example, by the chemical reactor fire at a Ciba Specialty Chemicals plant in Schweizerhalle in July 2001. This fire also produced a cloud of acidic smoke and revealed deficiencies in the emergency alert system. In such cases, reliable and necessarily rapid characterization and assessment of air pollution is still hampered by relatively poor data availability.
Fire site, soil, and groundwater
Remediation of the fire site and the contaminated soil took about 6 years, with 2,700 t of semi-combusted material being disposed of. For this purpose, Sandoz had to install a shed with a special off-gas system at the storage site. About 9 t of pesticides and 130 kg of organic mercury compounds had infiltrated the soil. The pollutants could be detected at depths of up to 11 m, although the top 6 m of soil was the most heavily contaminated (Munz and Bachmann 1993). Among the emergency measures instituted were groundwater lowering and sealing of the contaminated soil area, together with roofing and drainage installations. In a comprehensive risk analysis, the extent of the pollution was determined and the environmental behavior of the most important contaminants was assessed. The site was then decontaminated using a complex flotation process (Munz and Bachmann 1993).
The industrial site at Schweizerhalle was equipped with a sewer system that could be sealed off in the event of an oil spill. On the night of the fire, however, the seals were not closed, but even if the system had been sealed off, the fire-fighting water—between 10,000 and 15,000 m3—would still have made its way into the Rhine. The red-colored runoff was mainly discharged into the Rhine via a drain designed for uncontaminated cooling water (Fig. 1e). The runoff contained large quantities of active ingredients and auxiliaries from warehouse 956, together with combustion products. While these chemicals included non-toxic adjuvants such as the red rhodamine dye, most of the substances were toxic insecticides, herbicides, and fungicides. Sandoz’s initial claim that the Rhine was polluted with a harmless red dye, though certainly correct (see Fig. 1e), was unfortunately not the whole truth, as the dye in question was actually used to mark highly toxic mercury compounds. On the issue of fish mortality in the Rhine, see Fig. 1f and 2.
Figure 3 shows, for the most important pesticide substances, the quantities stored in warehouse 956 and the loads detected in the Rhine at the Village-Neuf monitoring station below Basel. It was possible to determine the latter values because weekly flow-proportional composite samples were continuously collected at Village-Neuf. This station participated in the Swiss National River Monitoring and Survey Programme (NADUF) and in the monitoring program of the International Commission for the Protection of the Rhine (ICPR). For almost all the substances, the loads measured in the Rhine were roughly proportional to the quantities that had been stored in warehouse 956. There were three exceptions—oxadixyl, dinitro-o-cresol (DNOC), and atrazine. The observation that substantial loads of oxadixyl were also found in subsequent samples from the Rhine suggested chronic contamination, i.e., that the substance was released both before and after the Schweizerhalle fire. This hypothesis was examined in cooperation with the chemical industry, and it subsequently proved possible to eliminate the source of contamination. Atrazine had not been stored in warehouse 956, but elevated loads of this herbicide were measured in the Rhine. The special situation with regard to atrazine arose from the fact that in October 1986 an accident had occurred at Ciba-Geigy, in which wastewater contaminated with atrazine had been retained in a basin. A few days before the Sandoz fire, Ciba-Geigy began to discharge this wastewater to the treatment plants, where the atrazine was only partly removed. On 31 October and the following days, approximately 0.4 t of atrazine was released into the Rhine from the Basel municipal wastewater treatment plant and the Pro Rheno industrial wastewater treatment plant. This explains the elevated loads of atrazine measured at Village-Neuf and at the other downstream monitoring stations on the Rhine. Initially, it was suspected that Ciba-Geigy might have acted as a “free rider”, taking this opportunity to dispose of the atrazine-laden wastewater. However, the sequence of events showed that this suspicion was unfounded. The company had, however, omitted to report the incident in accordance with the regulations.
The “Sandoz pollutant wave”—concurrent measurement of chemical and biological parameters and chemodynamic behavior of pollutants
Figure 4 shows the values measured for two chemical and two biological parameters at Bad Honnef (Rhine km 640) in North Rhine-Westphalia. The presence of the fluorescent marker rhodamine meant that the Sandoz pollutant wave could be relatively easily and clearly recognized. The total concentration of organophosphorus insecticides measured by gas chromatography rose to a maximum of 15 μg/l. The biological effect markers determined for the same samples (daphnia toxicity and cholinesterase inhibition) clearly indicated the insecticide impacts of the Sandoz wave. These findings represent one of the rare cases, in which measurements of chemical concentrations could be directly linked to observations of adverse biological impacts. In a certain sense, this can indeed be described as a textbook example.
The concentrations of disulfoton and thiometon measured at four monitoring stations along the Rhine are shown in Fig. 5, while Fig. 6 charts the disulfoton loads and includes a map of the Rhine indicating the location of the monitoring sites. On the basis of the known and in some cases field-measured substance data and the hydrological conditions in the Rhine, it was possible to model the Sandoz wave mathematically (Capel et al. 1988; Wanner et al. 1989). In Fig. 5b, the results of a model calculation performed for thiometon are plotted against the concentrations measured in the Rhine. It is evident from this comparison that the simulation of the Sandoz wave was successful. The most important factors taken into account to determine the environmental behavior were the relatively high water solubility (i.e., a weak tendency to sorb to suspended particles and sediments) and significant and quite rapid biodegradability. The latter property was responsible for the reduction in insecticide mass flows observed in the Rhine, as shown for disulfoton in Fig. 6.
Damage to the Rhine ecosystem
The German report on the Sandoz accident (DK-Rhein 1986) and subsequent publications (Güttinger and Stumm 1992) provide detailed accounts or overviews of the pollution and damage that occurred in the Rhine. In November 1986, the contamination of the Rhine resulting from the input of pollutants in fire-fighting water runoff had catastrophic effects on the river biota, including fish. Particularly striking was the eel kill, which spread from Schweizerhalle (Rhine km 159) to as far downstream as Rhine km 560 (Loreley near Koblenz).
However, other fish species were also severely affected, e.g., grayling, brown trout, pike, and pikeperch. Fish food organisms were acutely damaged in the vicinity of the accident site, while only weak effects were observed below Bad Honnef (Rhine km 640). Like fish, macroinvertebrates showed species-specific reactions to the toxic wave. Near Basel, the more sensitive Ephemeroptera and Trichoptera on the left bank of the river were affected, while Diptera and Gammaridae suffered less damage. In the Netherlands, effects of the Sandoz pollutants were noted especially in Tubificidae and Diptera larvae.
Within a few months, the recovery or recolonization of the Rhine benthos compensated for the obvious damage. That the damage was not even more extensive may be attributed to the fact that the Rhine is chronically exposed to significant chemical contamination and the channel was and remains heavily engineered; as a result, most sensitive species had already disappeared before the accident occurred.
Güttinger and Stumm sought to relate the damage observed to the ecotoxicity of the substances concerned. The eel population suffered acute toxicity at peak organophosphate concentrations of 10–20 μg/l (at Rhine km 500–600). According to available toxicity data, only endosulfan is acutely toxic in this concentration range. Unfortunately, however, LC50 values for eel were not available at that time, nor were measurements carried out of the actual concentrations of endosulfan in the Rhine. Dead grayling and brown trout were found almost as far downstream as Mainz. Here, the concentration of organophosphates persisting over a period of 96 h was around 5 μg/l, i.e., 1,000 times lower than the lethal concentration. In macroinvertebrates, adverse effects were observed at even lower concentrations. It is not possible to attribute the biological effects observed to specific concentrations of particular substances. However, it is clear that acute toxic effects occurred at significantly lower concentrations than the LC50 values would have suggested. It is necessary to consider not only synergistic effects of the substances concerned but also the fact that these pollutants added to the existing chronic contamination. In spite of dilution by tributaries, the concentrations of chronic pollutants tend to increase downstream in most cases, which may explain why Sandoz chemicals released in low concentrations still caused detrimental effects in the Netherlands. The ecosystem disturbance thus needs to be determined and evaluated in relation to chronic pollution and natural variations, although the natural reference conditions are not usually available.
Impacts on drinking water supplies and groundwater
As a precautionary measure, operations at drinking water utilities along the Rhine had to be temporarily suspended over a period of more than 18 days. In retrospect, it can be said that, given the abovementioned biodegradability, the Sandoz pollutants would have been eliminated by bank filtration and thus would not have entered treated drinking water supplies. The same applies to groundwater resources fed by bank filtrate from the Rhine. Also unaffected was the groundwater body and the associated artificial recharge system in the Hardwald area adjoining Schweizerhalle. This was essentially due to the favorable hydrological situation in the aquifer.