Grassland, woodland and crop fires

We are an independent AI lab researching real-world issues and solutions. The lab 'went live' just as the dry summer bit and wildfires began, so it started looking and asking questions. This led to a risk model and to FOI requests to West Midlands Fire Service and South Wales Fire & Rescue Service for clearer data to refine our wildfire-risk model (the where and the when). We have now analysed those first two datasets, and two findings have prompted us to extend this study to all UK fire and rescue services:


(1) Satellite fire-detection products (NASA FIRMS VIIRS/MODIS) detected only about 0.1% of the vegetation fires those services actually attended (41 of 5,379 in South Wales; 12 of 31,440 in the West Midlands). National vegetation-fire incidence can therefore only be measured from fire-service records - not from satellites.


(2) Fires recur intensely at a small number of places: in South Wales, half of all vegetation fires over five years occurred in roughly 150 one-kilometre cells. This makes a targeted, per-service 'dry-spell watch-list' genuinely feasible.


Our aim: with finer location-and-time data we can produce around 20 (or more) summer risk hotspots per service from the model, and share that map back with you. We are a self-funded lab (non-commercial research) looking to prove and improve our concept of AI-assisted tools.


I am requesting incident-level records of grassland, woodland, crop, and other vegetation/wildland fires - i.e. fires you classify under the "grassland, woodland and crop fires" category used in your returns to the Home Office Incident Recording System / MHCLG Fire and Rescue Data Platform - for the most recent 5 full financial years for which data is available, with the following fields per incident where held:


1. Date and time of call/attendance.

2. Location, at the finest real geographic precision you hold - ideally an OS grid reference or postcode;

failing that, the incident ground/street-level location description already

recorded in your Incident Recording System entry.

3. Estimated area affected (hectares, or your recorded size band), where held.

4. Motive/cause classification (accidental / deliberate / undetermined).

5. Whether the incident was recorded as a "major incident".


Why at this resolution: vegetation-fire counts by motive are already published at Fire-and-Rescue-Authority-area level. I am requesting only the finer-than-FRA-area location detail that the aggregate tables do not include, for small-area (LSOA-scale or finer) fire-risk research. This is not a repeat of data already public.


A spreadsheet (CSV or XLSX) is ideal. If any part would exceed the cost limit, please advise which fields or years are feasible within it rather than refusing the whole request. I need no personal data - only the incident attributes above.

Please see attached spreadsheet.

Data not held for question 5.

“Should you require the attachments mentioned above please contact customerinformationservice@lincolnshire.gov.uk with the below reference number and this can be provided.”