JG Border/line Support the author ↗

Borderline. About the project, sources and method

Recorded incidents / monthBar: records from that month. Selection: the chosen period.

Every tool lives here: full-text search and the filters the map does not carry. Filters apply to both views, so they are still there when you go back to the map. Click a row to open its card.

All words must match. Case, Polish diacritics and hyphens are ignored (Przewodów = przewodow, Ch-55 = Kh-55). Order: exact ID, place/model, aliases, description.

Country
Period
Type of event
Link to Russia

How firmly the event is tied to an actor. Grades A to D are explained on the about page.

Target
Object

Applies only to events where something flew.

Only
Register summary

Figures for the whole register. They do not respond to the filters above.

About the project

Borderline is a register of security incidents in NATO countries and Moldova linked to Russia's war against Ukraine: 10 types of event, from airspace violations to sabotage, arson and cyberattacks with physical effects. This is not a list of Russian attacks. We record separately what happened, and who named a perpetrator on what basis.

Why this map?

A single report says little on its own: a drone over an airport, debris in a forest, fighters scrambled. Placed together in time and space, the same incidents show where they repeat, which months see the numbers rise and how rarely any of it ends in a shoot-down.

The map does not argue on the reader's behalf. It keeps a date, a description and a source next to every point.

How to read Borderline?

Four views divide the work. The map: pick a period on the timeline, click a point, a card opens. The archive: full-text search and every filter, from type of event to effects. The story walks through selected cases, chapter by chapter.

One point is one event, placed at a location representative of the name given in the source rather than at an impact site. A circle with a number is a group of nearby events; clicking expands it. Some entries have no point and appear only in the archive.

Sources & method

The register holds 270 events from 24 February 2022 to 5 October 2026. Below: what it covers, what counts as one event, where the information comes from, how we assess responsibility and how precise the locations are.

Scope and exclusions

We observe events on the territory of current NATO members and Moldova. The research window opens on 1 January 2022; the earliest entry held dates from 24 February 2022. The first date says when we start looking, the second when we first have something. The absence of earlier entries is not a finding that nothing happened.

Other European states are out of scope, even where comparable events occurred. That is a limit of scope, not an assessment: the register is an editorial selection, not a complete census of incidents in Europe.

Some events happen outside any state's territory: in international waters and on the seabed. We do not assign such entries to the nearest landmass.

Individual cases are excluded by a scoping decision of the author and kept in a separate internal list. An exclusion asserts neither that responsibility has been settled nor that it has not.

The register also holds events with no Russian thread, and those stay too.

What counts as an event

One event is one entry, even when it carries several types or touches several countries. Types overlap: an airport disruption is often also an airspace violation. Per-type tallies are therefore higher than 270, by design.

We record separately whether the event occurred and whether the act was completed, attempted or foiled. A drone sighting is not evidence of an attempted attack, and a confirmed airport stoppage can have an unconfirmed cause.

Successive closures of the same airport are not added together twice, and the number of months in which something happened is not the number of events.

Verdicts, indictments and later findings are recorded as updates to an existing event. They do not create a new point on the timeline.

Sources and verification

The core of the register comes from the Polish and English Wikipedia survey lists, that is from secondary sources. 141 entries have since gained sources other than those lists: institutional statements, case documents and media reporting. 129 entries still rest on those two lists alone.

Verification is our editorial check, separate from what agencies or courts have established. An entry is marked verified when it rests on a source outside those two lists and that source was read in full. There are 140 such entries; 130 of 270 have not passed that check. Verified does not mean we have established who did it; unverified does not mean false, only unchecked against a source of our own.

For 60 entries the sources give different values, most often dates. The disagreement is recorded on the entry, together with the value we adopted.

Attribution and uncertainty

Officially attributed covers 21 entries, where a court or a competent state body has attributed it. This is not a count of proven attacks: an indictment is an allegation, not a verdict.

Suspected covers 19 entries: someone has named a perpetrator, but neither a court nor a state has done so formally. An official suspicion is still a suspicion.

Unresolved covers 59 cases examined and not settled, disputed, or where an attribution was rejected. A rejected attribution is not the same as a live suspicion.

Not assessed covers 171 entries with no assessment of responsibility in this register. It is the largest group. It does not mean "nobody has looked" or "no connection".

Every attribution carries the grade of its basis: A is a court finding, B an official attribution by a state body, C an official suspicion, D a documented journalistic investigation. D never masquerades as C.

The country where equipment was made, the nationality of the person who carried something out, who flew the device and who ordered the act are four different pieces of information. An object described as Russian-made is not a suspicion against anyone: such entries stay in the "unresolved" band, and the validator refuses any attribution resting on equipment origin alone.

Location and routes

Coordinates are usually derived from a place name rather than read from a source; 0 have been verified. A marker's outline shows the attribution band, not the precision of the coordinates. We deliberately do not draw an uncertainty radius, because there is no basis on which to estimate one.

The country a record is filed under is not the same thing as where the event happened in law. 9 events happened outside that country's territory: 8 in its exclusive economic zone, 1 in international waters. For 4 entries the sources do not settle it. The card says so plainly. A violation over the territorial sea counts as an event on territory, because that is how the law counts it.

32 entries have no point on the map: the event runs for tens of kilometres, or touches several places at once. Those entries live in the archive and say so, rather than being given an invented pin.

39 entries carry a route: 39 are schematic, straight segments between places named in the reporting, and 0 have documented geometry. A schematic route is not a radar track, and the path between the points is unknown.

Jakub Górnicki.

A Polish journalist and artist, winner of the Polityka Passport award, co-founder of Outriders. Borderline is another project in which reporting takes the form of a tool, after the reporting game Bottleneck about the Strait of Hormuz and after Zondatracker.

Bottleneck ↗ Zondatracker ↗ More projects: jakubgornicki.com/pl ↗

Open to corrections.

This is a working editorial prototype, not a finished publication. The data is incomplete and will keep growing.

If you spot an error in a date, a place or a source attribution, or you know of an incident that is missing here, get in touch. A message on LinkedIn or Instagram reaches me fastest.

LinkedIn ↗ Instagram ↗