Conflict layers use Uppsala Conflict Data Program records. The global register is built from conflict-year observations; candidate-event records provide dated places and fatality estimates where loaded.
| Layer | Unit | Loaded coverage | Use |
|---|---|---|---|
| UCDP/PRIO Armed Conflict Dataset | Conflict-year | 1946-2025 | Conflict identity, type, incompatibility, sides, location, intensity |
| UCDP Candidate Events | Dated event | 2025-August 2026 snapshot | Current events, place, actors, low/best/high fatalities |
What map counts mean
Conflict-map color counts distinct loaded conflict records intersecting the selected window and state. It is not territorial control, military effectiveness, force size, spending, or a fatality rate. Candidate-event markers are provisional records at the source boundary.
Candidate scope and casualty estimates
Each candidate conflict network is one UCDP-coded conflict, not a comprehensive regional casualty ledger. Candidate rows may mix incidents, period aggregates, and revisions, while the loaded source has no record-class field. The atlas therefore keeps low/best/high estimates on individual observations and does not add them into country, location, conflict, or Life and Death totals. GraphML downloads expose this boundary, included conflict and dyad identifiers, status, geometry eligibility, observed end, display end, and export date as first-class fields.
Parties and relationships
Side A and Side B reproduce source-side coding within a conflict. Same-side links are conflict-specific observed relationships, not permanent alliances. Opposing-side links do not imply an enduring diplomatic category. The network preserves this distinction.