Where the
Mourne fires
begin.
A satellite-based study of ten years of wildfire activity in the Mourne Mountains finds a pattern that is anything but random. The fires happen at the same times of year, on the same days of week, and — increasingly — in the same places.
fire detections
identified
in March or April
on a weekend
A decade of fire,
seen from orbit.
Every active fire over the Mournes for the last ten years has been recorded by NASA's FIRMS satellites. The map below renders that record as a heat density — bright zones are places where fires have been detected repeatedly across the decade. Cooler areas are where the satellites have rarely or never caught a thermal signal.
The pattern is not uniform. A small number of zones account for a disproportionate share of activity — and those zones do not correspond to remote interior peaks. They sit close to roads, gates, and forestry tracks. Visual precision has been deliberately blurred here to avoid implicating any specific access point publicly.
"Fires don't choose their starting points at random. They start where someone can stop, light, and leave."
A calendar
that tells everything.
Filter the dataset to the events most consistent with deliberate ignition — recurring locations, night-time starts, multi-year repetition — and the timing pattern is unmistakable. Every single one falls inside a six-week window that closes just after the legal heather-burning deadline of 14 April.
By month — suspected arson events
By day of week — suspected arson events
The Heather and Grass Burning Regulations (Northern Ireland) 2007 permit controlled vegetation burning only between 1 September and 14 April. This means every fire after 14 April is, by definition, illegal — regardless of intent. The clustering of events against this deadline suggests opportunism: a narrow window of receptive fuel, dry weather, and shrinking lawful cover.
Open data,
reproducible findings.
The analysis uses only freely available data and standard open-source tools. Anyone with a NASA Earthdata account, a laptop, and a few hours can reproduce the dataset from raw satellite detections to ranked recurring zones. The exact pipeline:
Pull
NASA FIRMS Area API queried for the AONB bounding box across MODIS, VIIRS S-NPP, and VIIRS NOAA-20 sensors. Ten years, three satellites, ~9,400 raw detections.
Deduplicate
Overlapping satellite passes produce repeat detections of the same fire. Records rounded to ~110m and 1-hour windows are collapsed, keeping the highest-FRP entry.
Cluster
DBSCAN with haversine distance metric — 500m radius, minimum 5 detections. Isolates locations where fire has recurred across years from one-off agricultural burns.
Classify
Each cluster scored on night-fraction, recurrence interval, fire radiative power, and seasonality. Cultural bonfire patterns flagged and excluded from arson analysis.
The full Python pipeline and intermediate data files are available on request to investigators, journalists, and statutory bodies. Open methodology is the basis on which findings should be evaluated and challenged. Errors found will be corrected publicly.
Some recurring
fires are cultural.
Three clusters in the dataset showed recurrence patterns initially consistent with the arson signature — same place, multiple years, late-night ignition. Closer inspection of the calendar dates revealed something different: every detection at these clusters fell on or around 12 July, between 01:00 and 03:00 UTC, with Fire Radiative Power values consistently below 3 megawatts.
These are cultural bonfires — primarily Eleventh Night community gatherings — with a smaller secondary signal on 31 December. They are recurring fires, but they are not arson. They are flagged here so that the temporal-recurrence methodology does not generate false leads, and so that investigators can confidently disregard them.
A short note on heather and grass ecology is also relevant. Gorse and heather take roughly 3–7 years to recover to their previous fuel density after a major burn. This means a deliberately lit landscape will not reburn at the same spot for years — forcing repeat offenders to move their ignitions to adjacent unburnt ground. The displacement pattern this creates is itself a forensic signature.
A summary you
can share.
A condensed one-page summary of the public findings is available below as a PDF. It contains the headline results, the methodology summary, and the bonfire-exclusion note — suitable for sharing with community groups, local councillors, or anyone interested in the topic.
Detailed coordinates of recurring ignition zones, per-event origin points, and predictive analysis are not included in the public release. That material is held back to protect the operational value of any subsequent investigation. Statutory bodies and investigative journalists may request access to the restricted version separately.
Download the one-page summary