The images looked like a major fire. What has been officially confirmed, however, is a failure of the external power supply that placed a heavier load on the safety flare system. Why the flame may become visible again – and what the air and bathing-water data actually show.
What many people could see from around Kvarner Bay
On the afternoon of 21 August, a dense, dark plume of smoke rose above the Rijeka refinery at Urinj. At the same time, the flare flame was visibly higher. From Kraljevica, Kostrena and parts of the island of Krk, it looked from a distance like a major fire in the industrial complex.
The images were real, but the first interpretation was not automatically correct. The INA statement published by the Municipality of Kostrena confirms neither a refinery fire nor an explosion. Nor does it confirm a direct lightning strike on a specific transformer station. What is substantiated is a different sequence: a thunderstorm, a failure in the external power grid, effects on the production units and a heavy load on the flare system.
The officially confirmed sequence of events
At around 4 pm, during the thunderstorm, the external power supply to which the refinery is connected temporarily failed. According to INA, this affected the production units and the safety flare system. The visible results were a higher flame and heavier smoke.
The power supply was subsequently restored. Stabilisation of the process system and the gradual restart of the units then began in accordance with prescribed safety procedures. This second stage matters: the end of the power cut did not instantly return the plant to normal operation.
Power failure, flare and air readings
Infographic showing the sequence from an external power failure through safe relief of the units to a larger flare flame and subsequent stabilisation
Why a refinery has a flare in the first place
A refinery flare is not a decorative chimney but a safety route for process gases. When units fail unexpectedly, shut down or restart, gases must be released from the affected systems in a controlled way. The flare burns them in a designated location instead of leaving them uncontrolled inside the installations.
That explains why a larger flame during a disruption can be part of the safety system. It does not mean that all visible smoke is insignificant. Combustion produces real emissions, and black smoke shows that combustion was temporarily not completely soot-free. The accurate conclusion is therefore: a safety function, yes – merely a visual effect, no.
The planned start-up of the DCU was the background, not the confirmed cause
The refinery had already announced on 20 August that it would start up the DCU unit between 20 and 23 August, with additional noise and smoke possible. DCU stands for Delayed Coker Unit, a facility that further processes heavy residues. The scheduled start-up explains why parts of the site were already in a sensitive operating phase.
Officially, however, the heavy smoke on Friday afternoon was linked to the unplanned failure of the external power supply. The two must not be merged. The unit start-up had been announced; the power cut during the thunderstorm had not. Only the combination of operating status, power interruption and subsequent stabilisation explains why the images looked so exceptional.
What the Urinj air-monitoring station showed – and what it does not prove
The public monitoring page of the county health institute displays unvalidated hourly readings for Urinj. For 4 pm it showed, among other values, 39.6 micrograms of nitrogen dioxide, 27.4 micrograms of sulphur dioxide and 9.2 micrograms of PM10 per cubic metre. The page did not mark an exceedance for the displayed pollutants during that hour.
This is not a blanket all-clear for the entire smoke plume. The readings have not yet been validated, one station cannot capture every location beneath a moving plume, and a missing reading is not a zero reading. Nor can a one-hour PM10 value be compared directly with a daily limit. The only reliable statement is that the station did not mark an exceedance for the displayed pollutants – not that there was no air pollution anywhere.
No new bathing warning – existing advice remains separate
At the time checked, Kostrena had not published a new bathing warning as a consequence of the power failure. There is likewise no official evidence of new marine pollution caused by the event. Conversely, a survey from 4 August obviously cannot assess what happened later.
Local advice not to bathe at Klančić and Podkvarovo near Urinj had already existed in the Kostrena area. Those notices have an older, separate history and must not retrospectively be presented as a result of the smoke on 21 August. Anyone wishing to swim there should check the specific monitoring site and the latest municipal publication, not just a general map of the coast.
Why the flare may grow larger again in the coming weeks
INA announced that a higher flare could temporarily recur during stabilisation and restart in the coming weeks. A visible flame after 21 August would therefore not automatically mean a new accident. But it would not justify dismissing every observation without checking it either.
The decisive factors are the time, the official notice and accompanying signs. A known, announced flare phase must be assessed differently from a suddenly expanding smoke plume with a bang, fire outside the flare, visible emergency crews or an alert message. New photographs alone prove neither a new emergency nor that only the previous operating condition is continuing.
What visitors can sensibly do if smoke appears again
Anyone who sees an unusual flare should first check the Municipality of Kostrena homepage and the regional air-monitoring page. If there is a strong smell or irritation, it is sensible to close windows temporarily and postpone strenuous outdoor activity until the plume has moved away or an official assessment is available. People with respiratory or cardiovascular conditions should follow their individual medical advice.
The 112 number is for a specific emergency – for example, fire outside the designated installation, injured people, severe acute symptoms or immediate danger. For the simple question of why a known flare is visible, municipal notices, operator information and monitoring data are better sources. Social-media images can help identify an event, but they provide neither its cause nor a health assessment.
What remains unknown
A final validated assessment of air quality during the event remains outstanding. The public statement also gives neither a detailed technical fault analysis of the external grid failure nor a complete emissions inventory. The information available therefore supports neither the conclusion that the event was pollutant-free nor that it was an uncontrolled major industrial disaster.
The confirmed core is narrower and more useful: the unusual smoke was real; an electricity failure and the resulting load on the flare system are officially confirmed. Power and the process system were stabilised, a larger flare may temporarily return during restart, and no new evacuation, health or bathing warning had been published by 22 August.
Evidence
Sources & date
Sources support the facts. Planning notes and recommendations are Golden Beach editorial assessments.
- Temporary power interruption caused increased smoke at the Urinj oil refineryMunicipality of Kostrena with INA statement · 21 August 2026; latest municipal information checked 22 August 2026external power failure, load on the flare system, restoration, stabilisation and a possible larger flame in the following weeks
- Notice from the Urinj oil refineryMunicipality of Kostrena / Rijeka refinery · 20 August 2026announced start-up of the DCU between 20 and 23 August and possible additional noise or smoke
- Air qualityTeaching Institute of Public Health of Primorje-Gorski Kotar County · unvalidated hourly readings from 21 August 2026; checked again 22 August 2026Urinj monitoring station, displayed hourly readings and limits of their interpretation
- Fifth assessment of sea quality in the Kostrena areaMunicipality of Kostrena · survey on 4 August, published 19 August 2026bathing-water status before the refinery event and distinction from existing local no-bathing advice
- Onečišćenje zraka BakarŽiga / Wikimedia Commons · photographed 2005; licence checked 22 August 2026lead and social image, CC0; historic archive photograph, not the incident of 21 August 2026
- Urinj 251208Roberta F. / Wikimedia Commons · photographed 25 December 2008; licence checked 22 August 2026inline image of the refinery and its location on Kvarner Bay, CC BY-SA 3.0; archive photograph
