How smoke from fires in Bosnia clouded the air from Split to Dubrovnik

On Sunday 6 September 2026, Split, its neighbouring towns and the islands woke up in smoke. The harbour master's office warned shipping of reduced visibility, which at sea had fallen to no more than 100 metres. The smoke came from a wildfire that had broken out on the Friday near Kupres in Bosnia and Herzegovina and spread towards Tomislavgrad and Prozor-Rama.^[2] The same day, the Teaching Institute of Public Health of Split-Dalmatia County reported that PM10 and PM2.5 levels in Split, Trogir and on the islands were several times higher than usual.^[1]

At the Split 2 station, the hourly AQI for PM2.5 climbed to 176 at 11 am, in the "unhealthy" band, with a PM2.5 concentration of 105.6 µg/m³. It is the highest reading in the month covered by the station's chart, from late August to mid-September.^[8]

Ivan Kovačević, the county's fire commander, told the Hina news agency that the wind, which had strengthened on Saturday evening, was carrying the smoke over the coast from Split to Pelješac, and described it as a "white gloom". Croatia's Civil Protection Directorate issued a notice to residents and tourists because the county 112 emergency centres were taking a growing number of calls, and the Croatian Firefighting Association asked people not to ring the fire service about the smoke.^[2]

The institute advised people to avoid strenuous outdoor exercise, with particular care for children, older people, pregnant women and anyone with chronic respiratory or cardiovascular disease.^[1] It put air quality at between 2 and 3 on a scale of 1 to 6 that it took from EU recommendations, which meant the air was considered loaded with particulate matter.^[1] Nenad Periš, who heads the institute's air-quality testing department, described the values as elevated in a statement issued through the county civil protection headquarters, but said they should not cause problems for healthy people.^[2] That scale is not the AQI shown on the AirQuality.city map, so the two figures cannot be compared directly.

A day later the picture had shifted. Periš said particle levels in Split were falling, while Mato Lakić of the Dubrovnik-Neretva County public health institute described the air around Dubrovnik as "quite poor" and changing from day to day with the wind and the fire. According to HRT, the wind direction had spread the smoke over hundreds of kilometres within a few hours. The broadcaster's chief meteorologist, Zoran Vakula, said it had been brought to the Adriatic by north-easterly flow, the bora.^[3]

Wind aloft and wind at the surface

Vakula also gave the reason the smoke lingered. The maestral that rose in the afternoon was not strong enough and blew only near the ground, not through the deeper layers of the atmosphere, so it failed to push the smoke back.^[3] Once smoke has been lofted, winds at different heights decide where it goes, and how much of it people breathe depends on whether it sinks into the layer of air near the ground. NASA described exactly this during the Canadian wildfires in July 2026: where the smoke stayed high, the effect on surface air quality was negligible; where it drifted lower, conditions worsened.^[5]

Pressure systems, upper-level winds and passing fronts all shape a plume's path. In NASA's analysis of 14 to 17 July 2026, a high-pressure system over the central United States positioned the jet stream so that smoke from Ontario could travel into the north-eastern US. About two days later a frontal system brought it over densely populated areas, and the sinking cold air behind the front held it near the surface.^[6] On its own, distance from a fire is therefore a poor guide to smoke exposure. A town under the main path of a smoky air mass can have worse air than one closer to the fire but off that path.

What PM2.5 shows, and what it does not

PM2.5, particles with an aerodynamic diameter of 2.5 micrometres or less, is the measure most closely watched during smoke episodes. According to the World Health Organization, it makes up about 90 per cent of the total particle mass emitted by wildfires. Smoke also carries PM10 and ultrafine particles, ozone, nitrogen and sulphur dioxide, polycyclic aromatic hydrocarbons, toxic metals such as lead, and microplastics, with the mix depending on what is burning. The WHO adds that smoke can travel thousands of kilometres and affect air quality and health far from the fire.^[4]

A sudden rise in PM2.5 at one station does not, however, prove that a particular fire is to blame. Fine particles have other sources, from traffic to home heating, and the number alone says nothing about where they came from. A stronger case rests on several lines of evidence: whether levels rise in sequence at stations along the path of the air mass, whether the airflow lines up with active fires, and whether satellite images show a plume on that track.^[5][6] In Dalmatia several of those pieces fitted together. Satellites were picking up numerous active hotspots at the fire in Bosnia and Herzegovina, the flow was north-easterly, from the mainland towards the sea, and civil protection and firefighters linked the smoke to that fire.^[2][3]

Canadian smoke over Washington

The same process played out on a far larger scale in North America in July 2026. By mid-July almost 850 fires were burning across Canada, more than 180 of them in Ontario. On 14 July the NOAA-21 satellite captured smoke being carried over southern Ontario, parts of Quebec and the US Midwest and Northeast. In many places the sky turned grey and yellow and the Sun orange, and air quality in Toronto reached unhealthy levels.^[5]

A NASA visualisation built from the GEOS atmospheric model showed the smoke travelling thousands of kilometres between 14 and 17 July and blanketing much of the United States. On 17 July the daily average PM2.5 concentration in the Baltimore-Washington area approached 200 µg/m³, prompting Code Purple air-quality alerts. The smoke was still there the next morning, until severe thunderstorms washed the particles out of the atmosphere.^[6] The US Environmental Protection Agency issued a statement the same day saying that a large part of the country was experiencing smoke from the Canadian fires, and pointed people to live air-quality data.^[7]

Split after 6 September

For Dalmatia, HRT reported on 6 September that the fire would keep burning and that only rain or a change in wind direction could bring relief. A day later it said the forecast change in the weather should provide it.^[2][3]

The Split 2 chart shows that the air did not clear all at once. Around midday on 7 September the hourly PM2.5 AQI jumped again, to 122. It stayed above 50, where the "moderate" band begins, almost without a break until the evening of 9 September, fell below that line overnight and through the morning of 10 September, then climbed back above it until the morning of 11 September. Throughout 12 September it stayed between 7 and 23.^[8] How much of that was smoke from Bosnia and Herzegovina and how much came from other sources is something the chart alone cannot show.