Nikolaos KamakiotisWildfire Risk Analyst
WithNERO, European network on extreme fire behaviourCroatian Meteorological and Hydrological Service (DHMZ)
FlEvo

Pisak

21 June 2025
Reconstructed by Nikolaos Kamakiotis, FlEvo Ltd

The 71 reconstructed isochrones of the Pisak fire, drawn in time order from the ignition point.
  1. 06:30Ignition above Pisak
  2. 07:20–07:30Peak spread, 55 m/min
  3. 08:00Head at the ridge crest, 119 ha
  4. 09:20The Bura fails
  5. 18:10Last isochrone, 411.6 ha

Ninety minutes in which a Bura-driven fire ran 3.7 km along the Dalmatian coast at up to 55 m/min, rebuilt from fire brigade cameras, firefighters' photographs, social media, radio logs, weather stations and satellite data, and replayed ten minutes at a time on the terrain where it happened.

Fire reconstructionNikolaos Kamakiotis, FlEvo Ltd
Meteorological analysisIvana Čavlina Tomašević, Jasmina Dobranić, DHMZ
Field data and radio logsMarko Maleničić, Omiš Fire Department
Supported byNERO COST Action CA22164

At 06:30 on 21 June 2025, fire broke out above the coastal road in Pisak, between Omiš and Makarska. Driven by a Bura wind off the Dinaric ridge, it ran west along the slope, crossed the road within 45 minutes and burned through Pisak, Marušići and Medići towards Mimice. Canadair aircraft could not work in the turbulence. By evening 412 hectares had burned.

The fire was reconstructed by Nikolaos Kamakiotis (FlEvo Ltd) from the evidence it left behind: videos and photos posted on social media, fire brigade camera frames, firefighters' photographs, 967 radio messages and satellite imagery. The meteorological analysis was carried out by the Croatian Meteorological and Hydrological Service (DHMZ), and field data came from the Omiš Fire Department, under the NERO COST Action.

Each coloured band on the map is ten minutes of fire spread: 71 isochrones from 06:30 to 18:10. As the clock runs, the records that carry their own timestamp appear one at a time, at the minute they were taken. Click any image to enlarge it.

Times are local summer time (CEST, UTC+2). The 3D map loads satellite imagery and terrain online.

Loading satellite terrain…
The setting
06:30
CEST · 21 June 2025
06:30Fire arrival time (CEST)18:10
Central Dalmatia

Pisak lies on the narrow coastal strip between Omiš and Makarska. Two DHMZ automatic weather stations bracket it: Split-Marjan, about 35 km to the north-west, and Makarska, about 19 km to the south-east. Fire brigade monitoring cameras at Nerežišća on Brač, Brela, Cista Provo and Vela Straža watched the coast.

412 hectares

The fire burned a strip 6.1 km long and up to 1.4 km deep, from the ridge down to the sea, through Pisak, Marušići and Medići towards Mimice. Unburned houses, plots and roads inside it are cut out of the burned area.

A ridge cut by gullies

The slope rises from the sea to over 600 m in little more than a kilometre and is cut by gullies running from the crest to the coast (red). East of Pisak the crest drops from over 600 m to about 300 m: the lowest gap in this section of the ridge.

1834 · Habsburg cadastre

The Habsburg cadastral survey of 1834 maps the slope as farmland: vineyards, olive groves and pasture around the villages, georeferenced here over today's terrain.

1834 · Vineyards

Vineyards (magenta) covered about 125 ha, 93 ha of them inside what would become the 2025 burn scar.

1968 · Scattered stands

Aerial imagery from 1968 shows natural tree vegetation confined to about 25 ha of scattered stands, 20 ha of them within the later scar.

2025 · Continuous fuel

By 2025 natural tree vegetation, largely Aleppo pine, covered 167 ha, 144 ha of it inside the burn scar: seven times the 1968 cover. Cultivation had shrunk to 15 ha of olive groves and a few small vineyards. Abandoned terraces had become continuous fuel from the ridge down to the houses.

Earlier fires

The slope had burned before. Fires in 2014 (3.6 ha, above Pisak) and 2018 (14.2 ha, mostly inside the later scar) burned patches of it. A 207 ha fire in 2015 burned immediately to the west (orange); the 2025 fire stopped about 400 m short of it.

Bura from inland to the sea

After a hot, dry June (Fire Weather Index at Makarska 44.7 on the day), cold, dry air over the continental interior spilled south-westward over the Dinaric coastal ridge and down to the Adriatic overnight (blue streamlines). The Bura peaked at Makarska at 03:50: 16 m/s, gusts 29.4 m/s. At ignition, 06:30, it was still 13.7 m/s with gusts of 26.5 m/s, and relative humidity was close to 12 %.

The direction hardly changed for more than nine hours: NNE to NE at Makarska, NE at Split-Marjan, until the Bura collapsed at 09:20.

Streamlines are schematic. Wind roses and dials: DHMZ automatic stations, 10-minute data; dials point the way the wind blows. Synoptic analysis: DHMZ.
06:30 · Ignition

Fire starts above state road DC8 in Pisak. It is reported at 06:36; crews from Omiš, Brela and Baška Voda are dispatched.

06:40 · Brela cameraTimestamp 06:40:31

Ten minutes in, the fire brigade camera at Brela, 7 km to the south-east, records smoke lying low along the slope above Pisak.

07:15 · Brela cameraTimestamp 07:15:22

The head has turned west along the slope, parallel to the coast. At 07:15 fire crosses DC8 in Pisak and burns between the houses; the road is closed a minute later.

07:33 · Brela cameraTimestamp 07:33:25

The fastest ten minutes of the day: between 07:20 and 07:30 the head advances 552 m, 55 m/min. At 07:33 Brela shows open flame on the slope above Pisak.

07:40 · Firefighter photoTimestamp 07:40

A firefighter photographs DC8 in heavy smoke with flames on the slope. The phone record fixes the time: 21 June 2025, 07:40.

07:41 · Nerežišća cameraTimestamp 07:41:23

From 07:41 the fire brigade camera at Nerežišća on Brač, 22 km across the channel, records the slope every 20 to 40 seconds. Scroll slowly: the frames run with the clock.

07:43 · Two frontsTimestamp 07:43:04

Smoke is held in a horizontal band and pushed west. A dense column breaks upward on the right of the frame, then a second: the slope fire and a separate coastal tongue, burning as two fronts.

07:45 · Convective cellsTimestamp 07:45:08

Four to five convective cells rise at regular spacing across the burning slope. Above them the smoke flattens and spreads west.

07:46 · The ceiling breaksTimestamp 07:46:38

A single dominant column penetrates the smoke ceiling, narrow at its base and expanding above. Below it, smoke keeps moving west at low level.

07:51 · A second columnTimestamp 07:51:39

A second, larger column develops, further west than the first: the head has moved.

07:52 · Cista Provo cameraTimestamp 07:52:07

From Cista Provo, inland to the north, the plume rises over the ridge at the same minute.

08:00 · Zagon cameraTimestamp 08:00:18

The Zagon camera sees the plume streaming out over the sea to the west. By 08:00 the head has run 3.7 km in 90 minutes and reached the ridge crest; 119 ha have burned.

08:10 · Nerežišća cameraTimestamp 08:10

The Nerežišća frames place the head near Medići. Residents and tourists in Marušići have been evacuated in heavy smoke since 08:02.

08:13 · Security camera, MedićiTimestamp 08:12:59

The key record of the day. A private security camera in Medići, running on its own clock, shows smoke closing in at 08:13 and flames entering the yard within a minute. It fixes the arrival of the front at a known point to the second.

08:20 · Nerežišća cameraTimestamp 08:20

From Nerežišća the front is now above Mimice. At 08:21 turbulence keeps the Canadair from approaching.

08:35 · Water drop, Brela cameraTimestamp 08:35:39

The Brela camera records a water drop on the front above Pisak. By 08:46 turbulence forces the Canadairs to withdraw to Divulje.

08:40 · Nerežišća cameraTimestamp 08:40

The front reaches Mimice. The head has slowed to 14 m/min, against 50 m/min an hour earlier.

09:07 · Last camera frameTimestamp 09:07:14

The last Nerežišća frame. Smoke lies along the whole coast; the head is moving at under 10 m/min. 242 ha have burned.

09:20 · The Bura gives way

At Makarska the north-easterly collapses and turns into a light south-westerly. The run is over: from here the fire spreads slowly, by flanks and reactivations. 269 ha have burned.

10:00 · Three-quarters burned

316 ha, 77 % of the final area, burned in the first three and a half hours.

12:00 · Slow flanks

The western edge creeps towards Lokva Rogoznica at 1 to 2 m/min. Reactivations keep reaching houses in Medići. 376 ha.

18:10 · Final perimeter

The last isochrone closes the burned area at 411.6 ha.

5.6 km in eight hours

Measured along the path of the head (white), the fire covered 3.7 km in its first 90 minutes, at 35 to 55 m/min, and reached the ridge crest by 08:00. After that it slowed to under 20 m/min, and after the Bura failed at 09:20 to a few metres a minute.

Rate of spread between successive isochrone crossings of the spread path.

1 · Oblique deflection

The Bura met the ridge at an angle and was deflected west, parallel to the sea front. That is why the burned area is a long blade along the coast, not a fan down to the sea.

The “unexpected” finding

2 · Up the gullies, against the wind

With the Bura blowing down the slope, the fire still ran up the gullies to the crest, ahead of the main front. The gullies acted as chimneys and left the finger-shaped northern edge of the burned area.

3 · Rotors on the lee slope

Below the crest the Bura broke into rolling eddies. Where they touched the ground the fire surged, in regularly spaced cells.

4 · A lid over the slope

A stable layer near 500 m held the smoke and heat down and pushed the fire sideways. Columns broke through it several times.

Seen from Brač, 07:44–07:49

The Nerežišća camera frames, looping: smoke held under a flat ceiling, then a column breaking through it.

5 · The gap east of Pisak

East of Pisak the ridge drops from over 600 m to about 300 m. The Bura was funnelled through this gap, faster and drier, onto the slope above the village.

06:30 – 14:20

How fast it moved

The rate of spread is measured along the path the head of the fire actually took, from the ignition point westward to the far end of the scar. The path is cut at each isochrone it crosses; the distance between successive crossings, over ten minutes, gives the rate. The head ran at 35 to 55 m/min for the first 90 minutes, peaking at 55 m/min (552 m) between 07:20 and 07:30.

3.7 kmAlong the spread path, 06:30–08:00
119 haBurned by 08:00, 29 % of the total
77 %Of the final area burned by 10:00
Rate of spread along the spread path
Rate of spread per 10-minute interval (m/min, bars) and cumulative distance along the spread path (km, line), CEST
≥ 40 m/min15–40 m/min< 15 m/minCumulative distance
Burned area over time
Cumulative area inside each isochrone (ha), CEST. Colours follow the isochrone map.

The head slowed sharply once it reached the ridge crest around 08:00. At Makarska the Bura recovered to 8–10 m/s around 08:00–08:30, collapsed after 09:00 and turned south-westerly at 09:20.

06:48 – 11:58

Seen from space

The new Meteosat Third Generation imager scans Europe every ten minutes. EUMETSAT's Land Surface Analysis service derives from it the Fire Radiative Power (FRP) of each fire pixel: the rate at which the fire radiates energy. Summed over the Pisak pixels and integrated in time, it gives the Fire Radiative Energy (FRE) released.

06:48First detection, 18 min after ignition
2.9 GWPeak FRP, 07:58, at the height of the run
9.9 TJRadiative energy, 06:48–11:58
Fire Radiative Power and fire growth
Total FRP of the Pisak pixels per 10-minute scan (MW, bars) and area burned per 10 minutes from the isochrones (ha, line), CEST
FRPArea burned per 10 min
Cumulative Fire Radiative Energy
Time integral of FRP (TJ), CEST

The satellite record and the reconstruction tell the same story independently. Radiative power climbs through the run and peaks at 2.9 GW at 07:58, exactly when the isochrones show the fastest growth of the day: 33 to 34 ha every ten minutes between 07:50 and 08:10. It falls away once the Bura fails, with a last pulse at 09:58.

Meteosat Third Generation FCI, FRP-PIXEL product (EUMETSAT LSA SAF, LSA-509).

Night of 20–21 June 2025

Meteorological analysis

The synoptic and mesoscale analysis was carried out by DHMZ (Ivana Čavlina Tomašević and Jasmina Dobranić) from observations, upper-air data and the ALADIN-HR model. It identifies a downslope Bura with a low-level jet of about 22 m/s near 500 m and very dry descending air over the coastal range, a weaker and shorter-lived version of the pattern behind the 2017 Split wildfire. The reconstruction uses the measured station winds throughout.

The evidence

Every evidence point on the map is a place and a time someone recorded. Each item was located from what it shows, matched to pre-fire imagery, and timed from its own clock where it has one, or placed by cross-checking against everything else.

Social media
TikTok, Facebook and YouTube videos and photos published by news websites (telegram.hr, Slobodna Dalmacija readers). Each is located from what it shows and drawn as a fire-front line where the flames are visible.
Firefighters' material
Photographs and video from the fire line, shared by the Omiš Fire Department, including a photo whose phone record fixes 07:40, and a private security camera stamped 08:12:59.
Radio log
967 radio messages from the Omiš Volunteer Fire Brigade (DVD Omiš), with positions, fire behaviour and unit movements.
Fire brigade cameras
Nerežišća on Brač (152 frames, 07:41–09:07), Brela, Cista Provo, Vela Straža and Zagon. Each frame is located, timed to the second and matched to the front position.
Field work
Site inspection with the Omiš Fire Station; interviews with local firefighters on flame heights, spotting and turbulence.
Weather
10-minute wind data from Split-Marjan and Makarska; ALADIN-HR model fields, profiles and cross-sections; synoptic and upper-air analysis.
Satellite
Pre- and post-fire Sentinel-2 imagery for the burned area; perimeters of past fires; Meteosat Third Generation fire radiative power every ten minutes.
Fuel history
Habsburg cadastral survey (1834), aerial imagery (1968) and recent imagery (2014, 2018, 2025), mapped polygon by polygon.
24–25 March 2026

The mission behind the reconstruction

This reconstruction was made possible by a Short-Term Scientific Mission funded by the NERO COST Action CA22164: Integration of Advanced Meteorological Analysis in High-fidelity Reconstruction of Wildland-Urban Interface Zone Wildfires Methodology: Reconstruction of Pisak wildfire, June 2025. It brought Nikolaos Kamakiotis to Croatia in March 2026. With a DHMZ agrometeorological officer and personnel of the Omiš Fire Station, the unit that managed the incident, he surveyed the burned area and held structured debriefings with the firefighters, refining through field validation the evidence already gathered from satellite imagery, social media and the press. The method was presented to the Split Fire Department and to the DHMZ synoptics team, and the agrometeorological officer was trained hands-on in the full reconstruction workflow.

The mission transferred the forensic reconstruction method developed by Nikolaos Kamakiotis to DHMZ, which contributed the synoptic meteorological analysis. Combined, the two allowed the identification of the sub-synoptic phenomena that appeared during the incident.

Nikolaos KamakiotisFlEvo Ltd, Cyprus · Wildfire Risk Analyst. Forensic fire reconstruction: evidence collection and location, milestones, isochrones, fuel history and spread analysis.
Ivana Čavlina Tomašević, PhDDHMZ · Synoptic and meteorological analysis
Jasmina Dobranić, MScDHMZ · Synoptic and meteorological analysis
Marko Maleničić, BScOmiš Fire Department, University of Split · Field data and radio logs
Working meeting at a café table with fire service colleagues
01 / 04Working session with fire service colleagues, 24 March 2026.
Group photograph with the firefighters of JVP Split in front of their engines
02 / 04After presenting the method to the Split Fire Department (JVP Split), 25 March 2026.
Walking down the steps to the sea through the burned area
03 / 04Surveying the burned area with the Omiš Fire Station, 24 March 2026.
The Pisak village limit sign on the coast road
04 / 04The end of Pisak on the coast road, 24 March 2026.
13–14 August 2026

Fourteen months later

Sentinel-2 false colour, before and after: the 2025 burn scar between Pisak and Lokva Rogoznica, then the 2026 fire from Lokva Rogoznica to the edge of Omiš. Animation: Nikolaos Kamakiotis.

On the evening of 13 August 2026, at about 20:21, a fire started beside the D8 coastal road at Lokva Rogoznica, where the 2025 fire had stopped. Driven by the Bura, it ran overnight along the same coast to Omiš, coming within about 50 m of the town centre.

It burned about 1,000 ha. Six people lost their lives and around 40 were injured. About 2,500 residents and tourists were evacuated, some of them by sea; houses, boats and more than a hundred vehicles were destroyed. Croatian foresters noted that this coastal strip has burned about ten times in 25 years.

Did we exploit the information from the 2025 incident as much as we could?

Could we have supported the authorities in reducing the impact of the 2026 incident?