Gjöll
Iceland's fire fatalities database, 1968–present
An open-access database documenting every confirmed fire fatality in Iceland since 1968. Triple-verified data and a finding that changed the conversation: no deaths in post-1998 buildings.
The Challenge
Iceland's official fire fatality statistics contained systemic inconsistencies. No single authoritative source existed that compiled, cross-referenced, and verified every fire-related death across the country's history. Policymakers, fire departments, and researchers lacked the reliable longitudinal data needed to assess whether building regulations and fire safety measures were actually saving lives.
The Approach
Manual data collection from multiple independent sources — news archives, fire department records, official statistics — with triple verification for every incident. The platform is built with Vue.js 3 and Vite on a Supabase backend, offering time-series analysis and downloadable datasets. All verified data are deposited in Gagnís, the University of Iceland's centralized research data repository, ensuring long-term institutional preservation.
Outcomes
Gjöll is a digital database and analytical platform that documents every confirmed fire-related death in Iceland from 1968 to the present. The project began in 2023 as a personal investigation into the reliability of official fire statistics — and quickly uncovered systemic inconsistencies in government records that had gone unquestioned for decades.
The most significant finding: no fire fatalities have occurred in buildings constructed after Iceland's modern building code came into effect in 1998. Over 27 years, not a single person has died in a fire in a code-compliant building. This is not an assumption — it is every verified record in the dataset.
The Name
Gjöll is the name of the river that flows through Helheim in Norse mythology — the realm of the dead. The name reflects both the gravity of the dataset and its purpose: to illuminate the darkest chapters of Iceland's fire safety history so that future tragedies can be prevented.
Methodology
Every incident is triple-verified against independent sources: news archives, fire department records, and official statistics. Where discrepancies exist, the verification chain is documented. The result is the most comprehensive and auditable record of fire fatalities in Icelandic history.
The platform is built with Vue.js 3, Vite, and Supabase. It provides time-series analysis tools, and downloadable datasets for researchers, fire departments, policymakers, and the public. All data are additionally deposited in Gagnís, the University of Iceland's centralized research repository.
Policy Implications
The data has direct implications for Icelandic fire safety policy. Current law permits fire safety inspections of commercial and industrial buildings, but not private residences — a regulatory gap that the dataset makes visible. The 2020 Bræðraborgarstígur fire in Reykjavik, which killed three people in the deadliest residential fire in recent Icelandic history, occurred in an older building that predated modern code requirements.
Building codes save lives. The data demonstrates this with 57 years of evidence. The question is what to do about the older housing stock where the risk concentrates.
Technology Stack
Lessons Learned
- Official statistics are not necessarily reliable. Cross-referencing revealed systemic gaps in government fire fatality records that had gone unquestioned for decades.
- Building codes save lives — the data shows it unambiguously. The 1998 building code eliminated fire fatalities in compliant structures over a 27-year period.
- Open data creates accountability. Publishing the full verified dataset means anyone can audit the methodology, challenge the conclusions, or extend the research.
Related Projects
View all projects →Beyond Fragmentation
Why AI governance fails by design — and why all harmful effects of technology are rooted in human governance.
Arctic Tracker
A data platform integrating 473,000+ CITES trade records, IUCN assessments, and illegal seizure data for 43 Arctic species. Co-authored preprint with Dr. Tom Barry now under review.