Skip to content

EM-DAT disaster impacts — introduction#

EM-DAT, the Emergency Events Database, is maintained by the Centre for Research on the Epidemiology of Disasters (CRED) at UCLouvain. It is the reference record of disaster impacts — how many people died, how many were affected, how much damage was done — for more than 26,000 events worldwide since 1900.

earthlens ships a single emdat backend covering the two routes that are actually programmatic. This page orients it. For the hands-on walkthrough see Usage; the rendered API is the Reference page.

The three EM-DAT products, and which one you get#

CRED publishes three things. They are not interchangeable, and the differences matter more than usual here.

Product What it is How earthlens reaches it
EM-DAT Public Data The living database, updated weekly Not used — the public.emdat.be portal is a manual XLSX download behind a login
EM-DAT Archive A validated snapshot of the public table, re-cut roughly yearly variables=["emdat:events"] — this backend, anonymously
HDX country profiles Yearly per-country summaries, not events The shipped hdx backend (see below)

The Archive is the one to use. CRED's own documentation calls it "the preferred choice for conducting research with EM-DAT": it is on the UCLouvain Dataverse, it follows FAIR principles, it has a DOI you can cite, and — unlike the portal — it needs no account. The trade-off is that it is a snapshot: if you need last week's events, the portal is still the only route, and it is a manual one.

Two sources, two shapes#

OUTPUT_KIND is per instance here, resolved from the dataset id you pass in variables:

  • emdat:events → tabular, a pandas.DataFrame. The impact record: one row per disaster, with deaths, affected, damage (nominal and CPI-adjusted), the country, and a coordinate pair.
  • gdis:points / gdis:polygons → vector, a pyramids FeatureCollection. GDIS (Geocoded Disasters) is a separate, openly-licensed derivative that geocodes EM-DAT's natural disasters onto GADM administrative units.

The two join on the disaster number: EM-DAT's DisNo. is 2009-0631-BGD, GDIS's disasterno is 2009-0631 — the same key with the ISO3 suffix dropped. The notebook does exactly that join.

Know these four things before you use it#

1. The two sources cover different periods and different hazards. The archive runs 1900 onwards and includes technological disasters (industrial and transport accidents) alongside natural ones. GDIS is frozen at 1960–2018 and is natural hazards only. A GDIS query will never return a 2020 flood, and never returns an industrial accident.

2. gdis:polygons is a 2.2 GB download. The GeoPackage of real admin-unit footprints is 2.2 GB compressed and 6.3 GB on disk. gdis:points is the same 39,953 locations as centroids in 1.09 MB — 2000× smaller — and it is the only GDIS distribution carrying year, latitude and longitude outright. Start with the points and reach for the polygons only when you genuinely need footprint geometry. The backend warns before it starts the large download.

3. The licences differ, and the archive's is restrictive. GDIS is CC-BY-4.0 — attribute it and you are done. The EM-DAT archive is CC-BY-NC-ND-4.0, and the Terms of Use go further than the licence tag suggests: free use is limited to academic organisations, universities, non-profit research institutions, international public organisations (UN agencies, multilateral banks, national governments) and media, for research, teaching or information purposes. Anything else is Commercial Use and needs a paid agreement with CRED. The terms also forbid redistributing the database or building a derivative database from it. earthlens therefore keeps no hidden cache and redistributes nothing: the archive workbook is fetched into your output directory and reused from there, the filtered result is written beside it, and a LicenseWarning naming these restrictions is raised on every emdat:events download.

4. Only GDIS needs credentials. emdat:events is anonymous. The GDIS granules used to be served openly from NASA SEDAC, but that host is gone and they now live in NASA Earthdata Cloud, so gdis:* needs an Earthdata Login — plus a one-time acceptance of the SEDAC data-use agreement, without which the download returns a 401. See Usage.

What this backend deliberately does not do#

  • Country summaries. The EM-DAT country profiles on the Humanitarian Data Exchange are already reachable with the shipped hdx backend and are not re-implemented here:

    from earthlens.core import EarthLens

EarthLens("hdx", hdx_id="emdat-country-profiles", path="out").download()

  • The GraphQL API. A GraphQL endpoint exists at https://api.emdat.be/v1 and answers with Missing API Key, please provide a value for the header: Authorization. It is not documented anywhere on the current EM-DAT documentation site, and there is no published way for a new user to obtain a key, so earthlens does not build on it. If CRED documents it again, it would slot in as a third source.

  • The XLSX portal. Superseded by the Archive for every programmatic purpose.

CRED's own external resources page lists two other geocoded derivatives that earthlens does not yet ship:

  • FLODIS — flood disasters 2000–2018, joining EM-DAT fatalities and damage plus IDMC displacement to satellite flood footprints from the Global Flood Database (Nature Sci Data 2023). Keyed on disasterno, so it joins straight onto what this backend returns.
  • Geo-Disasters — Teber et al. (2025), geocoding climate-related EM-DAT events.

Citation#

Cite both sources when you use them together:

  • EM-DAT, CRED / UCLouvain, Brussels, Belgium — https://www.emdat.be. Archive DOI 10.14428/DVN/I0LTPH.
  • Rosvold, E.L. & Buhaug, H. (2021). GDIS, a global dataset of geocoded disaster locations. Scientific Data 8,
  • NASA SEDAC, DOI 10.7927/61jv-th84.