Incident directory

1983 - Milford Haven




  • Explosions


Near miss


Date of event

30th August 1983     

Time of event

Approximately 10:53 (INITIAL CALL)

Name of premises

Amoco Refinery.


Milford Haven, Pembrokeshire, Wales.

Service area

Dyfed Fire Brigade (DFB), now Mid and West Wales Fire and Rescue Service (MAWW).

Nature of incident

Fire and subsequent boil-over

Property type

Oil storage tank (tank 11).

Premises use


Construction type and materials

Steel. 78 meters in diameter and 20 meters high with a capacity of 94,110 M3. A bund 90 meters wide and 180 meters long was also present.


Members of staff at refinery.

Fire source and location of fire



Brief Synopsis

Dyfed Fire Brigade were called to a fire incident at The Amoco Refinery, Milford Haven on the 30th August 1983 where an oil storage tank number 11 had caught fire. This tank contained over 46,000 tonnes of North Sea crude oil. The fire began at around 10:53 and developed to fully involve the storage tank. At approximately midnight the first boilover occurred. Several firefighters were injured during the firefighting operation. At least 6 firefighters were injured to varying degrees including an RAF Fireman (Dyfed County Fire Brigade, 1983).

Taken from date). Amoco refinery fire 1983.

The fire was first noticed by a works firefighter who alerted the Refinery Fire Brigade. They responded by sending its 105 foot Hydraulic Platform and it’s 4,500 gallon capacity Foam Tender. These were manned by 4 works firefighters. Within the next 3 hours there would be 150 firefighters manning 43 appliances. Before the stop message was sent there would be 70 appliances at the scene.

When the first refinery fire appliances arrived 4 lines of hose were got to work from hydrants along with a foam line from the Foam Tender to supply the refineries hydraulic platform which was got to work in the South East corner of the site, adjacent to the storage tanks bund wall. The boom on the HP was raised and foam was projected onto the oil tanks top which appeared to be now 50% engulfed in flame. The foam monitor on the HP was capable of supplying around 5,000 gallons per minute of foam onto the fire. Due to the fire not being seen to spread it was assumed that the foam had sealed the tank, but the tank was still well alight.

In the meantime a predetermined attendance consisting of 5 Pumps, an Emergency Tender and a Control Unit were mobilised from Dyfed County Fire Brigade. On arrival the message “Large Oil Tank alight” was sent and a “Make Pumps 10” message sent. Nine minutes later at 11.16 am a Divisional officer arrived and immediately sent the message “Make Pumps 15” he then sent an informative message informing control that a tank containing 56,000 gallons of crude oil was well alight, and 10 jets were now in use. Shortly afterward he sent another message requesting that large scale foam mobilising need to be started and that other refineries be requested to start their “Mutual Aid Scheme”. Pumps were then made to 20 with 5 Hydraulic Platforms required.

At 13.31 pm the Chief Fire Officer (CFO) took over and sent a message to say that a massive cooling operation was in force and that a foam attack wasn’t being used as they were waiting for other special appliances from other refineries to arrive before starting the attack. By 3.00 pm 26 Pumps and 7 foam tankers were at the scene along with other special appliances and 150 firefighters.

The fire was estimated to be consuming around 300 tonnes of crude oil per hour, so it was decided to draw of oil from the affected tank and several of those adjacent. It took until midnight to empty the adjacent tanks. Once empty, foam was introduced to these tanks.

It was calculated that a stock of 45,000 gallons of foam would be needed on site before a safe attack on the fire could be commenced. This foam would arrive in various commercial tankers, but when it arrived firefighters were confronted by these tankers having many different and many non-standard couplings. Firefighters on site fabricated adaptors from whatever they could find within the vast refinery, and eventually found means of preparing the discharge the foam into fire service tankers.

To fight the fire 40,000 gallons of a 6% foam mix per minute would need to be produced. A trial application of foam was attempted. This was achieved by using a roof monitor from an RAF Foam Tender that had arrived to assist firefighting. This test proved successful, so senior officers decided to assemble every single bit of foam making equipment on the site ready for a full on attack on the flames. Much of this equipment would be ready should oil spill into the protective bund wall and reduce the risk of this oil catching fire.

After further discussion between senior fire officers and refinery experts it was decided to leave the attack for at least 4 hours due to the oil being drawn from the affected tank still being cool. Also it meant more of the contents could be salvaged. It also gave time for more supplies of bulk foam to arrive. Suddenly, and without any warning a “Boil Over” occurred and many thousands of tonnes of oil flowed into the bund wall and caught fire. Suddenly the fire was to cover an area of about 4 acres. Firefighters ran for cover, many receiving burns to their hands and faces, while many suffered other injuries in the attempt to escape the severe heat. Appliances near to the tanks caught fire and many others had paintwork blistered by the intense heat. Immediately senior officers ensured a roll call was taken, amazingly all persons were accounted for. Ambulances took many of the firefighters for treatment at local hospitals.

Firefighters were then faced with the problem of lack of firefighting hose. The intense heat and fire by the tanks had melted much of the hose that had been laid in preparation for the attack, other hose had burst. Hose Layers with additional hose were ordered onto the incident. By now many of the original crews were exhausted and being relieved by fresh crews. These crews were warned of the potential of a further “Boil Over”. At 2.20 am another “Boil Over” occurred, firefighters were once more withdrawn, and much of the freshly laid out hose was damaged beyond use. Fires started to appear in the cladding of adjacent oil tanks. These tanks had foam applied to them and these fires were quickly extinguished. Fire crews worked quickly and efficiently, and within ½ hour the damaged hose was replaced and cooling jets replaced and got to work. These cooling jets were kept at work all night, and at 8.00 am it was decided to commence a full attack on the fire due to the potential risk of a huge “Boil Over”.

By now 67,000 gallons of bulk foam were available for use at the incident. The fist monitor was used to tackle the fire in the bund wall area in order to allow other foam monitors to be got to work as the area cooled. Soon 7 large foam monitors were at work from 3 fire fronts. By mid-afternoon the fire appeared to be diminishing in its intensity, even though the blanket of foam was being broken down due to the intense heat. By early evening the fire was under control. The fire in the bund wall was now out and a thick blanket of foam covered the area. Only small pockets of fire remained within the tank.

The application of foam was ordered to continue through the night, and at around 2.00 am on 1st September the fire suddenly flared back up as the foam blanket was broken. Senior Officers fearful of any sudden further flare up decided to use a mobile crane from within the refinery. To this they lashed a foam monitor to the jib. This allowed further foam to be applied to cool the fire and maintain the blanket of foam. This final attack was successful and the fire was extinguished. The application of foam was now continued for several hours in order to ensure the oil and metal of the tanks was cool and no further risks of fire be possible.

At 22.30 hour the Chief Officer sent the stop message informing control that no further appliances or firefighters would be needed.

The incident required 44 pumps. 30 of these pumps came from Dyfed, the remainder from surrounding brigades 14 foam tenders were used along with 66 commercial tankers (, unknown date).

Photo 1

Courtesy of Mid and West Wales Fire and Rescue Service.

Photo 2

Courtesy of Mid and West Wales Fire and Rescue Service from Dyfed County Fire Brigade, 1983.

Photo 3

Courtesy of Mid and West Wales Fire and Rescue Service.

Photo 4

Courtesy of Mid and West Wales Fire and Rescue Service.

Photo 5

Courtesy of Mid and West Wales Fire and Rescue Service.

Photo 6

Courtesy of Mid and West Wales Fire and Rescue Service.

Main findings, key lessons & areas for learning

Further information hoping to be identified and still to be located.

Fire & Rescue Service summary of main findings, conclusions, key lessons & recommendations

Taken from the Mid and West Wales Fire and Rescue Service Incident Report:



(a) Fire Protection Measures

(i) Fixed/Mobile Protection

Two documents are relevant:

The Home Office Model Code of Principles of Construction and Licensing Conditions (Part II) for Distributing Depots and Major Installations, 1968 and the Re fining Safety Code - Part 3 of the Institute of Petroleum Mode l Code of Sa fe Practice in the Petroleum Industry, 3rd Edition, 1981.

(a) Fire Protection Measures

(i) Fixed/Mobile Protection (Continued)

Both of the above documents specify tank spacing; bun ding and fire fighting measures which, by implication treat floating roof tanks as a lower risk than fixed roof tanks. Indeed the latter code, only two years old, recognises the otherwise good safety record of floating roof tanks by stating "with floating roof tanks, it is usual to design fire-fighting facilities to cater for rim fires only". Section 2.1.3. of the Home Office Code states, "(c) Water cooling can be achieved by fixed nozzles ........ Fixed nozzles are not necessary on floating roof tanks".

The Home Office code states that foam application may be by "mobile protection". Such "mobile protection" has obvious attractions to the Petroleum Industry, since some mobile equipment would be inevitable, irrespective of what fixed protection was provided. The Home Office code also states that "Fixed foam connections are not necessary on floating roof tanks".

As a result of the Code's guidance, and especially since there appeared no reason to view it otherwise, mobile protection only, consisting primarily of a 105 ft. hydraulic platform, a 4,500 gall. capacity Foam Tanker, and a Foam Tender, was provided.

(ii) Water Supplies

The Home Office Code states, "a fire main capable of supplying at all times a sufficient output of water at an adequate pressure should be provided". A 24" (610 mm) diameter public water main feeds a 2.5 million gal.(113,649,000 litre) fresh water tank situated adjacent to the Fire Station, off road 4. Three fixed fire pumps, each of 2,000 g.p.m. (9000 litre) rated output at 200 p.s.i. (13.8 bar) located near the Steam Plant draw water from the tank and supply a 10" (254 mm) diameter above ground fire main which serves the refinery site. The nearest hydrants to the incident were on roadways 5 and H, on the north and east of tank O 11. The Brigade used 2 x 1,000 g.p.m. (4,500 litre) pumps to supplement the fire pumps, using a facility provided to similarly draw water from the tank and pump into the fire main. There is also a diesel fir~ pump of 3,000 g.p.m. (13,500 litre) rated output at 200 p.s.i. (13.8 bar) which fed water from the raw water lagoons (46m gals.) into the fire main.

Owing to draw-off for the drenchers to L.P.G. spheres the water available for fire fighting and immediate-vicinity cooling was considerably reduced. Further draw-off was required to provide the water curtain to tanks 609 and 610. As a result, the fire main could only supply water to a maximum of three foam monitors and further water had to be provided by water relay from the lagoons some half-mile distant. In order to provide the amount of water for the scale of foam attack required, three water relays were provided; the first utilised Dyfed appliances having 1,000 g.p.m. pumps, the second 500 g.p.m. pumps and six 70 mm lines, were laid. The third relay utilised 6" hose from a hose layer, some fifteen appliances were committed to the supply of water to the fire ground.


(a) Fire Protection Measures (Continued)

(iii) Foam Compound Supplies

The Home Office Code states, "Foam stocks should be kept in an agreed location. The type and quantity should be arranged by consultation with the Licensing Authority". Since both the above Code and Industry Code view the greatest threat with floating roof tanks as rim fires, sufficient foam compound, immediately available for this purpose, would be reasonable. However, the level of foam stocks was considerably higher: Amoco Refinery held 7,500 gals. (34,095 litres) in two tanks, one located on the jetty, in addition to the 4,500 gals. (20,455 litres) within the foam tanker. Similar levels of foam compound were held by the other refineries; there was the Brigade's compound in two tankers, as well as on fire tugs on the Haven. In total, over 36,000 gals. (163,655 litres) were available. As a contingency, arrangements had been made for foam compound from the manufacturers - details of addresses and telephone numbers were immediately to hand - and from neighbouring Brigades.

The Avonmouth oil fire Bristol in 1951, involved tanks containing a total of nearly 4 million gals'. (18,184,000 litres), the largest of which was 9 5 ft. (29m) in diameter and 35 ft. (10.6m) high. Over 47,000 gals (213,615 litres) of foam compound were used.

The Duisburg oil fire, West Germany in 1979, involved 24 tanks ranging in capacity from 1,500 m3 to 4,700 m3. Over 50 tonnes of foam compound were used.

The Milford Haven oil fire involved a single tank 78 metres in diameter, 20 metres high of 94,110 m3 capacity and a bund 90 metres wide by 180 metres long. At the time of the fire they contained 60,000 m3 46,798 tonnes of crude oil.

167,899 gals. (763,055 litres) of foam compound were used. In calculating foam compound requirements several sets of figures were available.

The Manual of Firemanship, Part 6b, quote s a formula based on No. 10 (or equivalent) foam making branches.

Using the formula which Esso Petroleum provide in their Fire Manual, however, seemed more realistic and this suggested that some 41,383.5 gals. compound would be required just for the tank.

The final foam usage proves that this latter formula is far more accurate and this incident, along with the previous incidents mentioned, shows. the table and formulae in the Manuals of Firemanship to be seriously deficient.

(b) Siting and Bunding of Tanks

The Codes referred to above, allow up to three floating roof tanks of the capacity of that involved in the fire, in a single bund. Tanks need be no more than 0.3 times the diameter of the larger tank, but not less than 10 m a part.


(b) Siting and Bunding of Tanks (Continued)

In this case, a rectangular single-tank bund (180 m x 90 m) had been created, which was approximately 115% of the tank capacity, with the tank constructed within some 10 m of the east end.

In an adjoining bund, on the other side of road 5, about 61 m distant, were tanks 609 and 610. These two tanks required cooling water curtains from a very early stage and in the event, still suffered damage. It is a reasonable assessment that, had there been three tanks of the size of O 11, in a single bund, in accordance with the Codes, or had tanks 609 and 610 been closer, they would have become heavily involved in the fire, with quite considerable consequences for the eventual outcome.

Several questions are therefore raised as a result of this incident and taking into account previous incidents, some of which have been briefly mentioned earlier:

1.  Should the Codes of Practice require fixed protection to all tanks, by product storage rather than design?

2.  Should such fixed protection be wholly or partially automatic for more rapid response?

3.  In order to minimise still further the risk from radiated heat and thereby improve the chances of successful firefighting, should the Codes require greater separating distances and individual bunds for tanks over a certain capacity?

4.  In view of the almost indefinite shelf life of modern foam compounds is it now financially feasible to require larger stocks to be maintained immediately available?

5.  Whilst the siting of flare stacks in relation to risk areas such as tankage, appears generous, and this incident may well be unique, in the light of current economic practices of low flaring levels, should there be a requirement for more frequent cleaning of flare stacks? Or indeed, should not some alternative form of emergency pressure relief be developed, which does not require flaring?

6.  As with the fire in France in 1959, maintenance of the floating roof played a significant part in the rapid development of the fire. Should therefore, tank maintenance be given a higher priority and should there be a requirement that, when faults are discovered, immediate steps are taken, rather than waiting until the tank is emptied and made ready for routine inspection and maintenance?

7.  Foam compound requirement formulae, have been devised, based on tests which at the most, use pre-burn times of less than ten minutes. Not only this fire, but others have shown that, when pre-burn times are of greater extent, larger quantities of foam_ are required, particularly as a result of greater break-down due to high oil temperatures.

8.  In addition, the greater surface areas which may need to be blanketed, seem to require a greater foam application than an extrapolation of formulae derived from application to small surface areas would suggest.

9.  Thermal updraughts are created in such large fires and the sheer size, both height and diameter, severely affect the predicted and otherwise efficient rate of application.

There is a dear need to review currently accepted formulae for calculating foam compound requirements and tests should take into account more realistic pre-burn times.

In view of the intense heat generated and .sustained for Jong periods, should some form of shielding, either fixed or mobile, be provided for firefighters around hydrants and monitor positions?


FBU summary of main findings, conclusions, key lessons & recommendations

Further information hoping to be identified and still to be located.

Other report summary of main findings, conclusions, key lessons & recommendations

Further information hoping to be identified and still to be located.

IFE Commentary & lessons if applicable

None produced at this time.

Known available source documents

Further information hoping to be identified and still to be located.

FRS Incident Report/s

No information identified to date and/or still to be located.

FBU Incident Report/s

No information identified to date and/or still to be located.

Health & Safety Executive (HSE) Incident Report/s and/or improvement notices

No information identified to date and/or still to be located.

Dyfed-Powys Police Incident Report/s

No information identified to date and/or still to be located.

Welsh Ambulance Service Incident Report

No information identified to date and/or still to be located.

Building Research Establishment (BRE) Reports/investigations/research

No information identified to date and/or still to be located.  

Coroner’s report/s and/or regulation 28 notices

Not relevant.

Other information sources

Further information hoping to be identified and still to be located. (unknown date). Amoco refinery fire 1983. [online]. Available at [Accessed 20th September 2016].

French Ministry of Environment. (2008). Boilover of a crude oil tank 30 August 1983 Milford Haven (Wales) United Kingdom. [online]. Available at [Accessed 20th September 2016].

Dyfed County Fire Brigade. (1983). Report of the investigation into the fire at Amoco refinery 30th August 1983. Dyfed.

Service learning material

No information identified to date and/or still to be located.

Videos available

Milford Haven Boilover – August 30, 1983.  Youtube. (2013).  [Accessed 20th September 2016].

(In German) Youtube. (2010). Feuerwehr: Boilover / Boil-over bei Tankbränden.wmv. [online]. Available at [Accessed 14th May 2017]. 

Go back