Skip to content

IFRC DREF

The IFRC Disaster Relief Emergency Fund (DREF) provides immediate financial support to National Red Cross and Red Crescent Societies for early action and response to disasters and crises.

Collection Metadata

Data Sourcing

API Endpoints

  • Base URL: https://goadmin-stage.ifrc.org/api/v2/
  • Events Endpoint: /event/
  • Parameters:
    • dtype: Disaster type filter
    • appeal_type: Appeal type filter — GO's atype is 0 (DREF) or 1 (Emergency Appeal)
    • id: Specific event ID

Data Retrieval Process

  1. Events are filtered based on:
  2. Appeal type — an event is kept when every one of its appeals is atype 0 or 1, so the collection covers DREF and Emergency Appeal operations despite its name
  3. Valid disaster type

An event with no field reports ("field_reports": []) still passes this filter and still produces an event item — it just produces no impact items (see below).

  1. For each event:
  2. Basic event information is extracted
  3. Field reports are used to generate impact items
  4. Country information is used for geometry generation, unioning polygons across all listed countries when there is more than one

Data Structure

Event Data Model

{
    "id": int,
    "name": str,
    "summary": str,
    "dtype": {
        "name": str  # Disaster type
    },
    "countries": [
        {
            "name": str,
            "iso3": str
        }
    ],
    "disaster_start_date": datetime,
    "appeals": [
        {
            "atype": int  # Appeal type: 0 = DREF, 1 = Emergency Appeal
        }
    ],
    "field_reports": [
        {
            "id": int,  # disambiguates impact item IDs across reports
            "countries": [
                {
                    "name": str,
                    "iso3": str
                }
            ],
            "num_dead": int,
            "gov_num_dead": int,
            "other_num_dead": int,
            "num_injured": int,
            "gov_num_injured": int,
            "other_num_injured": int,
            # ... similar pattern for other impact metrics
        }
    ]
}

Accepted Disaster Types

GO exposes 24 disaster types (dtype.name); Monty ingests the 15 below, matched on the exact string (including the Pluvial/ prefix, which is part of the GO vocabulary):

  • Earthquake
  • Cyclone
  • Volcanic Eruption
  • Tsunami
  • Flood
  • Cold Wave
  • Fire
  • Heat Wave
  • Drought
  • Storm Surge
  • Landslide
  • Pluvial/Flash Flood
  • Epidemic
  • Civil Unrest
  • Insect Infestation

The other 9 are out of scope, each for a reason checked against GO's own appeal counts and event names rather than assumed — full analysis in #96:

GO disaster type DREF + EA appeals Why it's excluded
Other 542 Not a hazard by definition — there is nothing to resolve
Population Movement 404 An impact (displacement), not a hazard
Food Insecurity 128 An impact, usually of drought (MH0401)
Complex Emergency 37 Not one hazard: sampled events mix heat waves, missile strikes, armed conflict and displacement under the same dtype
Transport Accident 10 Mode-ambiguous: sampled events are road, rail, air and water accidents with no dominant mode to default to
Chemical Emergency 2 Category-ambiguous: sampled events are spills, explosions, fires and poisonings, not predominantly one HIP
Biological Emergency 1 Category-ambiguous: sampled events include measles and FMD outbreaks alongside water contamination and a poison-gas incident — overlaps Epidemic without being a subset of it
Radiological Emergency 0 No DREF/EA appeals to date; would map to TL0601 if populated
Transport Emergency 0 No DREF/EA appeals to date; a duplicate of Transport Accident's ambiguity

Counts are DREF + Emergency Appeal operations, checked 2026-07-31. This list is a decision, not a filter of convenience — a type only stays excluded because either it isn't a hazard or its dtype genuinely spans more than one HIP with no defensible default, not because mapping it was inconvenient.

Item Mapping

Event Items

STAC Field IFRC Field Required Notes
id ifrcevent-event-{id} Yes Prefixed ID
geometry Generated Yes From country ISO3, unioned across all of the event's countries when there is more than one
datetime disaster_start_date Yes Start date of the disaster
title name Yes Event name
description summary No Event summary
start_datetime disaster_start_date Yes
end_datetime disaster_start_date Yes Same as start (no end date provided)

Monty Extension Fields

Field Source Notes
monty:episode_number Fixed value (1) DREF doesn't track episodes
monty:hazard_codes dtype.name Mapped to standard hazard codes
monty:country_codes countries[].iso3 Array of ISO3 codes
monty:src_event_id Source event ID Unique identifier of the event

Impact Items

Impact items are generated from every field report on the event (not just the first), with multiple impact types per report:

Impact Type Source Fields Category
Death num_dead, gov_num_dead, other_num_dead ALL_PEOPLE
Displaced num_displaced, gov_num_displaced, other_num_displaced ALL_PEOPLE
Injured num_injured, gov_num_injured, other_num_injured ALL_PEOPLE
Missing num_missing, gov_num_missing, other_num_missing ALL_PEOPLE
Affected num_affected, gov_num_affected, other_num_affected ALL_PEOPLE
Assisted num_assisted, gov_num_assisted, other_num_assisted ALL_PEOPLE
Potentially Affected num_potentially_affected, gov_num_potentially_affected, other_num_potentially_affected ALL_PEOPLE
Highest Risk num_highest_risk, gov_num_highest_risk, other_num_highest_risk ALL_PEOPLE

Each field report yields one item per impact type that has at least one truthy (non-zero, non-empty) source field — a field report where all three source fields for a type are explicit zeros (e.g. num_dead: 0) produces no item for that type, with: - ID format: ifrcevent-impact-{event_id}-{impact_type}-{field_report_id} — the field report ID disambiguates items when more than one report on the same event carries a value for the same impact type. - Geometry generated from the field report's own countries (unioned when there is more than one), falling back to the parent event's geometry when the field report lists no usable country. - monty:country_codes taken from the field report's own countries[].iso3, falling back to the parent event's country codes when the field report lists none - Impact details including: - category (ALL_PEOPLE) - type (specific impact type) - value (first truthy value from the three source fields, in the order shown in the Source Fields column above — the other two are discarded, not cross-checked) - estimate_type (PRIMARY)

Hazard Type Mapping

GO carries a disaster type name (dtype.name) and nothing more granular, so the mapping below is keyed on that exact string. The 2025 UNDRR-ISC code is the reference classification for the Monty extension; the GLIDE and EM-DAT codes are its counterparts from the same row of the cross-classification mapping and are never chosen independently of it.

IFRC dtype.name GLIDE EM-DAT UNDRR-ISC 2025 (Reference) Cluster Notes
Earthquake EQ nat-geo-ear-gro GH0101 GEO-SEIS All earthquake phenomena were consolidated into GH0101 in 2025
Cyclone TC nat-met-sto-tro MH0306 MH-WIND Depression or Cyclone — the default, see the cyclone rule below
Volcanic Eruption VO nat-geo-vol-vol GH0201 GEO-VOLC Eruption, phenomenon unspecified. HIP 2025 has no volcanic chapeau (GH0201–GH0205 only), so GH0201 carries the general case, per the VO / nat-geo-vol-vol crosswalk row — not because the event is a lava flow. Refine to GH0202 (ash/tephra fall) or GH0204 (lahars) only when the operation names that phenomenon
Tsunami TS nat-geo-ear-tsu MH0705 MH-MARINE Reclassified from Geological to Meteorological & Hydrological in 2025
Flood FL nat-hyd-flo-flo MH0600 MH-WATER Flooding (chapeau)
Cold Wave CW nat-met-ext-col MH0502 MH-TEMP
Fire WF nat-cli-wil-wil EN0205 ENV-FOREST Wildfires — the default, see the fire rule below
Heat Wave HT nat-met-ext-hea MH0501 MH-TEMP
Drought DR nat-cli-dro-dro MH0401 MH-PRECIP
Storm Surge SS nat-met-sto-sur MH0703 MH-MARINE
Landslide LS nat-geo-mmd-lan GH0300 GEO-GFAIL Gravitational Mass Movement (chapeau), matching the GDACS/EM-DAT/GLIDE convention
Pluvial/Flash Flood FF nat-hyd-flo-fla MH0603 MH-WATER Flash Flooding
Epidemic EP nat-bio-epi-dis BI0101 BIO-INFECT General infectious disease. BIO-INFECT has no chapeau either, so BI0101 carries the unspecified case per the EP / nat-bio-epi-dis crosswalk row, despite its "Airborne Diseases" label. Refine when the operation names the pathogen: cholera → BI0110, viral haemorrhagic fevers → BI0109, parasitic → BI0105
Civil Unrest — — SO0103 SOC-CONF The cross-classification mapping has no GLIDE or EM-DAT row for the Societal hazard type, so this triplet is UNDRR-only — a single code is a valid monty:hazard_codes value
Insect Infestation IN nat-bio-inf-inf BI0401 BIO-INSECT General infestation. nat-bio-inf-inf is used rather than nat-bio-inf-loc, whose crosswalk row is inconsistent (it appears twice, once labelled "Insect pest infestation"→BI0401 and once "Locust infestation"→BI0402); refine to BI0402 only when the operation names locusts specifically
The cyclone rule

GO's Cyclone type is not exclusively tropical: sampling 200 of the 476 Cyclone operations on GO (2026-07), 1 is explicitly extratropical (ARG: Extratropical Cyclone - Misiones) and 137 explicitly name a tropical system (hurricane, typhoon, tropical cyclone/storm, or the TC/TCs abbreviation, e.g. TC Alfred, TCs Nika (Toraji) and Ofel) in the event name. The remaining 62 are ambiguous from the name alone — often a bare storm name (Cyclone Beryl, Cyclone Gamane) that is almost certainly tropical but isn't textually marked as such.

So: default to MH0306 / TC / nat-met-sto-tro (the crosswalk's "Depression or Cyclone" row), and refine using the event name only:

  • extratropical / extra-tropical → MH0307 / EC / nat-met-sto-ext
  • hurricane, typhoon, tropical, or \bTCs?\b → MH0309 / TC / nat-met-sto-tro

Both refinements keep TC / nat-met-sto-tro as the GLIDE/EM-DAT companions where the crosswalk pairs them with MH0306 or MH0309 — that pair is not tropical-specific; the same TC / nat-met-sto-tro row resolves to MH0306, MH0308 or MH0309 depending on the UNDRR name, so keeping it on the default row does not overclaim precision the way the UNDRR code would. Only the extratropical branch changes the GLIDE/EM-DAT pair, because that's a genuinely different crosswalk row (EC / nat-met-sto-ext).

This under-covers the ambiguous 62 — a bare storm name is not detectable without cross-referencing IBTrACS or another tropical-cyclone track database, which is out of scope for a string match on name — but it never makes an outcome worse than today's blanket MH0306: every event either gains a more specific code or keeps the default it already had. The repo-wide MH0306 vs MH0309 convention for sources that don't disambiguate at all is still being settled in #94; this rule only changes what IFRC DREF does with information GO already gives it.

The fire rule

GO has a single Fire type covering both vegetation and structural fires, and wildfires dominate it — of the 25 most recent, the Greek, French, Spanish, Algerian, Moroccan, Tunisian, Portuguese, Chilean, Bolivian, Patagonian and Victorian operations are all wild or forest fires, against four structural ones. Mapping the whole type to the industrial-fire code would file most DREF fire operations under the Technological family and stop them correlating with the EN0205 / WF that EM-DAT, CEMS, IDMC, IDU, GLIDE, DesInventar and PDC use for the same fires.

So: default to EN0205 / WF / nat-cli-wil-wil, and use TL0305 / FR instead when the event name — not summary — identifies a structural or industrial fire:

  • factory, industrial, plant, refinery, warehouse → TL0305 / FR / tec-ind-fir-fir (Fire, Industrial Failure)
  • landfill, market, building, structural, residential, apartment, camp, slum → TL0305 / FR / tec-mis-fir-fir (Fire, Miscellaneous Failure) — matching the crosswalk's own split between the two EM-DAT fire keys

summary is deliberately excluded. GO's summary is narrative impact prose, and it uses exactly these words to describe what a wildfire destroyed — "threatening residential areas" (Greece, Turkey wildfires), "urban-forest interface" (Patagonia), damage tallies that count "buildings" (a Korean wildfire that damaged over 200 of them) — not what kind of fire it was. Checked against every DREF Fire operation on GO (225, 2026-07): matching name and summary produces 15 false positives, all of them genuine wildfires reclassified as structural because their damage narrative mentions a building or a neighbourhood; restricting the match to name alone produces zero false positives against the same set, because operations that are actually structural or industrial say so in the title (Factory Fire in Daejeon, Nairobi Residential Fire, AMSA Landfill Fire, Hargeisa Market Fire) rather than only in the body text. The rule is still a heuristic — GO gives no field beyond free text to key on — but it is now the version that survived being checked against the real data, not the first list that seemed reasonable.

Note

All three classification codes (GLIDE, EM-DAT, UNDRR-ISC 2025) should be included in the monty:hazard_codes array for maximum interoperability, and all three must come from the same crosswalk row — a triplet assembled from different rows resolves inconsistently downstream. More specific hazard codes can be used following the characteristics of the event, as the volcanic, epidemic and fire rows above describe.

This mapping enables standardized hazard categorization while preserving IFRC's original disaster type classification in the source properties.

Quality Control Notes

  1. Events are filtered to DREF and Emergency Appeal operations (atype 0 and 1).
  2. Only the 15 accepted disaster types are processed.
  3. Impact values are not cross-referenced between sources: for each type, the first truthy value among the primary/government/other source fields (in that order) is taken as-is, and the other two are silently discarded — there is no reconciliation or consistency check between them.
  4. Geometry is generated from country codes, unioning polygons across all listed countries when there is more than one; impact items fall back to the parent event's geometry whenever none of their own field report's countries resolves to a geometry (an empty list, or a list where every entry fails geocoding).
  5. Every field report on an event is processed independently — an event with N field reports and M impact types with a truthy value can produce up to N × M impact items, distinguished by the field report ID in each item's ID.