← Back to Frameworks

EM-DAT: The International Disaster Database

URL: https://www.emdat.be/ Type: Academic research database (University of Louvain / WHO) Active since: 1988

Mission

EM-DAT is the Emergency Events Database, created in 1988 as a joint initiative between the Centre for Research on the Epidemiology of Disasters (CRED) at the University of Louvain (UCLouvain) in Brussels, Belgium, and the World Health Organization (WHO). It exists to serve "the purposes of humanitarian action at national and international levels" by providing a systematic, validated record of the occurrence and impacts of mass disasters worldwide.

EM-DAT is fundamentally different from the other frameworks in this collection: it is not a risk assessment framework, a forecasting platform, or a policy advocacy organization. It is a historical database -- the most comprehensive single record of disaster events on Earth. Its value lies in the empirical foundation it provides for all other risk assessment and forecasting efforts.

What They Track

Disaster Types

EM-DAT classifies disasters into two main groups:

Natural disasters (following the 2014 IRDR Peril Classification and Hazard Glossary, up to four classification levels):

  • •Geophysical: Earthquakes, volcanic activity, mass movement (dry)
  • •Meteorological: Storms (tropical cyclones, extratropical storms, convective storms)
  • •Hydrological: Floods, landslides (wet), wave action
  • •Climatological: Droughts, glacial lake outbursts, wildfires
  • •Biological: Epidemics, insect infestations, animal incidents
  • •Extraterrestrial: Meteorite/asteroid impacts, space weather

Technological disasters:

  • •Transport accidents: Air, water, rail, road
  • •Industrial accidents: Chemical spills, explosions, structural collapses, mining disasters, radiation events
  • •Miscellaneous accidents: Fires, other

Data Fields

For each disaster event, EM-DAT records:

  • •Location: Country, region, specific areas affected
  • •Dates: Start and end dates
  • •Disaster type and subtype: Per the classification above
  • •Human impact: Deaths, injuries, people affected, people made homeless
  • •Economic impact: Estimated damage (USD), insured losses where available
  • •Response: Declaration of state of emergency, call for international assistance
  • •Sources: Which sources provided the data for each record

Inclusion Criteria

A disaster is included in EM-DAT if it meets at least one of the following thresholds:

  • •10 or more people killed
  • •100 or more people reported affected
  • •A declaration of a state of emergency
  • •A call for international assistance

Coverage

  • •Temporal: 1900 to present (though data quality and completeness improve significantly after 1960 and again after 1988 when systematic collection began)
  • •Geographic: Global -- all countries
  • •Volume: Over 27,000 mass disaster events recorded

Methodology

Data Collection

1. Daily monitoring: CRED staff monitor predefined sources (news agencies, UN agencies, NGO reports, government communications) daily for disaster events meeting inclusion criteria. 2. Source identification: For each event, staff search for additional sources beyond initial detection. Source selection is influenced by the reported figure itself, the area and period it covers, the chronology of the sources, and their reliability. 3. Source hierarchy: Preference is given to institutional or scientific sources over press sources. UN agencies, national disaster management agencies, and reinsurance companies are generally prioritized over media reports. 4. Manual validation: Data is manually processed from textual documents. There is no automated data ingestion -- every entry is reviewed by a human researcher. 5. Periodic thematic updates: In addition to daily monitoring, CRED conducts periodic systematic reviews of specific disaster types or regions to identify missing events and update existing records.

Data Quality Assurance

  • •Standardized methodology: Consistent classification and coding rules applied across all events and time periods
  • •Systematic validation: Cross-referencing multiple sources for each event
  • •Source transparency: The sources used for each data point are recorded, enabling independent verification
  • •Version control: The EM-DAT Archive is maintained on UCLouvain's Dataverse under FAIR principles (Findable, Accessible, Interoperable, Reusable)

Known Data Limitations (Acknowledged by CRED)

  • •Pre-1960 data is significantly less complete than modern records
  • •Economic damage estimates are highly variable in quality, depending on what national governments and international organizations report
  • •"People affected" is an imprecise metric with inconsistent definitions across countries and disaster types
  • •Slow-onset disasters (droughts, epidemics) are harder to delimit than sudden-onset events (earthquakes, floods)
  • •Disaster counts per year are partly a function of monitoring intensity, not just disaster frequency

Data Availability

EM-DAT provides multiple data access pathways:

  • •Public data portal: public.emdat.be -- free access for non-commercial use after registration. Allows custom queries by country, region, year, and disaster type.
  • •Excel download: Data can be exported as flat tables in Microsoft Excel format (.xlsx) via the "Access Data" tab or Toolbox after login.
  • •UCLouvain Dataverse: The EM-DAT Archive is available under a CC-BY-NC-ND license, hosted on UCLouvain's institutional repository. This is the recommended source for research use, adhering to FAIR principles.
  • •Humanitarian Data Exchange (HDX): Annual data summaries per country (country profiles) available on data.humdata.org.
  • •Our World in Data: EM-DAT data is visualized and made interactively explorable at ourworldindata.org/natural-disasters (with attribution).
  • •World Bank Data360: EM-DAT data accessible through the World Bank's data platform.
  • •Monty STAC Extension: EM-DAT integrated into the IFRC's Monty disaster data standard.
  • •No public API: As of 2026, there is no REST API for programmatic access. Data must be downloaded through the portal or Dataverse. (The ACAPS API includes some EM-DAT-derived data.)

License: CC-BY-NC-ND for the archive. Free non-commercial use; commercial use requires negotiation with CRED.

Strengths

  • •Unmatched historical depth: No other single database provides comparable global disaster data going back to 1900. The 27,000+ event records spanning 126 years constitute the empirical backbone of disaster research worldwide.
  • •Standardized classification: The IRDR-based disaster classification system enables consistent categorization across countries, time periods, and disaster types. This standardization is essential for trend analysis and cross-country comparison.
  • •Human-validated data: Every record is manually reviewed -- no automated scraping or algorithmic classification. This introduces human judgment but also catches errors that automated systems would miss.
  • •Source transparency: Recording which sources informed each data point enables external validation and quality assessment. This is a feature many databases lack.
  • •Institutional durability: Housed at a major European university with WHO partnership, EM-DAT has maintained continuous operation since 1988. This institutional stability matters for a database whose value compounds over time.
  • •Wide adoption: EM-DAT is the standard reference database for the IPCC, UNDRR, World Bank, IFRC, and hundreds of academic publications. This broad adoption creates network effects -- improvements benefit the entire disaster research community.
  • •Free non-commercial access: Researchers, NGOs, and governments can access the full database without payment.

Blind Spots

  • •Threshold bias: The inclusion criteria (10 deaths / 100 affected / state of emergency / international assistance call) systematically exclude smaller disasters. This creates a survivorship bias: only events large enough to cross the threshold are recorded. In aggregate, the unreported small disasters may affect more people than the large recorded ones.
  • •Pre-1960 data gaps: Data before 1960 is significantly incomplete, making long-term trend analysis unreliable for early decades. Claims about "increasing disaster frequency" based on EM-DAT must account for increasing monitoring capacity, not just increasing disaster occurrence.
  • •Economic damage data quality: Damage estimates vary enormously in methodology and reliability across countries. Wealthy countries with insurance industries report more accurate damage figures than low-income countries. This biases economic impact analysis toward wealthier nations.
  • •No risk assessment: EM-DAT records what *has happened*, not what *might happen*. It provides no forward-looking risk assessment, probability estimation, or early warning capability. It is a rearview mirror, not a windshield.
  • •No sub-national granularity in standard output: While events are geolocated, the standard data export is country-level. Sub-national analysis requires additional processing.
  • •Slow-onset disasters underrepresented: Droughts, famines, and ecological degradation unfold over months or years and are difficult to delimit as discrete "events." EM-DAT's event-based structure handles these poorly compared to sudden-onset disasters.
  • •Technological disasters sparse: The natural disaster records are far more complete than technological disaster records. Industrial accidents, chemical spills, and infrastructure failures are inconsistently captured, particularly in countries with limited reporting.
  • •No cascading or compound event analysis: EM-DAT records individual events but does not systematically track compound events (e.g., earthquake triggering tsunami triggering nuclear meltdown) or cascading impacts across sectors.
  • •No climate attribution: Events are classified by type but not by underlying driver. There is no systematic distinction between climate-change-driven events and non-climate-driven events within the same category.

Relevance to Ark Instinct

What we can use:

  • •EM-DAT is the empirical ground truth for Ark Instinct's disaster intelligence layer. Every claim about disaster trends, historical patterns, or baseline risk should be checked against EM-DAT records. It is the closest thing we have to a canonical record of what has actually happened.
  • •Specific integration opportunities:
  • •Historical base rates for disaster types (e.g., how many major floods per year, what is the death toll distribution for earthquakes) to calibrate forward-looking risk models.
  • •Country-level disaster profiles to complement INFORM's risk scores with actual disaster history.
  • •Trend analysis for climate-related disaster types to provide empirical grounding for climate risk assessments.
  • •Validation data: when INFORM or WEF predict high risk for a country/region, does EM-DAT confirm that disasters actually occur there at higher rates?
  • •The HDX and Dataverse access points enable automated or semi-automated data ingestion.

What we should avoid:

  • •Do not use raw EM-DAT counts to claim "disasters are increasing" without controlling for improved monitoring and reporting capacity over time. CRED themselves caution against this naive interpretation.
  • •Do not rely on EM-DAT for forward-looking risk assessment. It tells you what happened, not what will happen. For that, use INFORM, Metaculus, WEF, and domain-specific models.
  • •Do not use economic damage data for precise comparisons across countries or decades without adjusting for inflation, GDP changes, and reporting methodology differences.

Methodology to adopt:

  • •EM-DAT's manual validation with source transparency is a model for data quality in Ark Instinct. Every data point should be traceable to its source, and automated ingestion should be supplemented with human review.
  • •The IRDR disaster classification system is a well-established taxonomy that Ark Instinct should adopt rather than inventing its own.
  • •The CC-BY-NC-ND licensing model and FAIR data principles should guide how Ark Instinct structures its own data outputs.

Sources

  • •EM-DAT. "The International Disaster Database." https://www.emdat.be/ (accessed 2026-04-03)
  • •EM-DAT Documentation. "Data Accessibility." https://doc.emdat.be/docs/data-accessibility/ (accessed 2026-04-03)
  • •EM-DAT. "Publications." https://www.emdat.be/publications/ (accessed 2026-04-03)
  • •Humanitarian Data Exchange. "EM-DAT Country Profiles." https://data.humdata.org/dataset/emdat-country-profiles (accessed 2026-04-03)
  • •UCLouvain. "CRED: Epidemiology of disasters." https://www.uclouvain.be/en/research-institutes/irss/cred-epidemiology-of-disasters (accessed 2026-04-03)
  • •ScienceDirect. "EM-DAT: the Emergency Events Database." https://www.sciencedirect.com/science/article/pii/S2212420925003334 (accessed 2026-04-03)
  • •Wikipedia. "Centre for Research on the Epidemiology of Disasters." https://en.wikipedia.org/wiki/Centre_for_Research_on_the_Epidemiology_of_Disasters (accessed 2026-04-03)

Contributors

  • •entity: nous-aeternos

type: ai-agent role: author date: 2026-04-03