AAGE — Amhara Advocacy Group in Europe
Documentation, research and advocacy for civilian protection in Ethiopia's Amhara region.
AAGE — Amhara Advocacy Group in Europe
Documentation, research and advocacy for civilian protection in Ethiopia's Amhara region.
Sources, verification, limitations, and a script anyone can run to reproduce every figure on the Monitor.
The Amhara Drone Strike Monitor documents — incident by incident — the use of aerial weaponry against civilian populations in Ethiopia's Amhara region since August 2023. Where reporting is restricted, where journalists are denied entry, where survivors are silenced, the act of counting is itself a form of resistance.
This page describes how we count.
Each entry corresponds to one event in which aerial weaponry — drones or fixed-wing aircraft — was reported to cause civilian casualties or damage to civilian infrastructure in the Amhara region. For each incident, we record where available:
We do not document military-on-military engagements, accidents unrelated to aerial weaponry, or events outside the Amhara region.
As of 26 May 2026, the dataset contains:
| Year | Incidents | Killed | Injured |
|---|---|---|---|
| 2023 | 28 | 369 | 90 |
| 2024 | 91 | 694 | 192 |
| 2025 | 20 | 129 | 21 |
| 2026 (to April) | 8 | 0 | 0 |
| Total | 147 | 1,192 | 303 |
The eight 2026 incidents are recorded as having occurred but carry no casualty figures: information from the affected sites has not yet been independently confirmed, a recurring pattern under information blackout.
The dataset draws from three publications across two organisations. Two of the three — the foundational sources for the period August 2023 through May 2025 — are produced by the same organisation, the Amhara Association of America (AAA). This concentration is documented openly under Limitations.
AAA's strikes-specific publication. AAA cross-references local witness accounts, social media verification, satellite imagery analysis where available, and reporting from independent Ethiopian outlets.
AAA's broader periodic report series, with detailed drone-strike data for Amhara. The May 2024 – May 2025 edition records 136 drone and air strikes resulting in 860 civilian casualties (656 killed, 204 injured). These figures overlap substantially with AAA's strikes-specific tracker over the May – December 2024 window; our consolidation treats the strikes tracker as primary reference and AWU as corroborating expansion source for the overlap period.
The Amhara Advocacy Group in Europe maintains its own incident log, sourced from diaspora field contacts, Amharic-language reporting (Borkena, Addis Standard, ESAT, and others), and direct survivor testimony when accessible. This component fills the period October 2024 onward and is the source most actively undergoing independent verification — see Limitations.
Of 147 incidents in the dataset:
Secondary sources cited across the dataset, in descending frequency: OHCHR (4 incidents), Addis Standard (4), ACLED (3), BBC Amharic (3), ReliefWeb (2), The Reporter (2), plus single appearances of BBC, UNSC, the New Humanitarian, Addis Insight, Mail & Guardian, Borkena, ECFR and the ICRC.
The Monitor presents two complementary views of the same documentation effort:
The summary figures at the top of the Drone Monitor page reflect every incident in the dataset, with totals computed mathematically from per-incident records.
The embedded tracker (the live Drone Monitor) reflects the AAA-published dataset alone, frozen at the January 3, 2025 report. The narrower figures there (152+ strikes · 823+ killed · 253+ injured) represent AAA's officially published minimum.
The two views are not contradictory: the AAA tracker is the canonical AAA-documented subset; the consolidated dataset extends it with AAGE field documentation and AAA's subsequent Amhara War Updates series. The arithmetic gap between them — roughly 240 additional deaths documented for the August 2023 – December 2024 window in the consolidated dataset versus AAA's official minimum — is the AAGE contribution above AAA's published floor.
Each incident carries one of four levels:
That 88% of incidents are at Reported level reflects the information environment, not lower credibility. Under sustained telecommunications blackout, a single source attribution is often the maximum independent confirmation that can be obtained. Upgrading to Corroborated or Verified requires sources we cannot fabricate.
The dataset records seven target categories, taken directly from per-incident classification:
| Category | Incidents | % of dataset |
|---|---|---|
| Civilian — type unspecified | 82 | 55.8% |
| Schools | 30 | 20.4% |
| Residential | 13 | 8.8% |
| Transport | 11 | 7.5% |
| Livestock & livelihood | 5 | 3.4% |
| Health facilities | 4 | 2.7% |
| Markets | 2 | 1.4% |
| Total | 147 | 100% |
The largest category, Civilian — type unspecified, contains incidents where civilian casualties were confirmed but the type of struck site could not be ascertained from available sources. Its size measures not where the strikes did not hit, but where the documentation could not reach. As field corroboration improves, incidents may migrate from this category to a more specific one.
The Livestock & livelihood category records strikes against cattle and pastoral subsistence. While not strictly «civilian sites» in the humanitarian sense, the deliberate destruction of subsistence in a pastoral economy is documented as a recurring pattern with direct consequences for civilian survival.
We document this work openly, including what is not yet complete.
146 of 147 incidents reference AAA as a source; 128 reference AAA alone. The single-source pattern reflects an information environment where AAA is the most consistent and best-resourced documenter of strikes in the Amhara region. Broadening the documentary base — through independent corroboration from international monitors (OHCHR, ACLED, OCHA) and direct primary-source reporting — is our current priority for incidents from October 2024 onward.
Repeated and prolonged internet shutdowns in the Amhara region (most notably from August 2023 onward) routinely delay the documentation of incidents by weeks or months. Some incidents from the affected periods may not yet be recorded.
In several recent incidents, survivors have declined to speak to documenters out of fear of retaliation. These incidents are recorded with the Reported tag and minimal detail.
The embedded AAA tracker reflects the dataset as of January 3, 2025 and is not updated by AAGE. Updates depend on AAA's own publication schedule. We do not modify the AAA tracker; the consolidated figures shown above the tracker reflect AAGE's own ongoing compilation.
Every total displayed on the Monitor is computed from the underlying data file. The dataset is publicly available as JSON:
A verification script (Python, no dependencies) is included in the site repository at scripts/verify-dataset.py. Running it reproduces the headline figures and the per-category breakdown shown on the Monitor:
$ python3 scripts/verify-dataset.py Total incidents: 147 Sum of killed: 1192 Sum of injured: 303 Earliest date: 2023-08-12 Latest date: 2026-04-20 [full output below]
Any reader — journalist, researcher, or sceptic — can re-run this in under five seconds and verify that the page's figures match the data exactly. If a discrepancy is ever found, it is a bug: please report it.
If you believe an incident is mis-documented, missing, or wrongly categorised:
Contact: [email protected]
For confidential tips: Signal / PGP channel in preparation — see Contact.
We accept corrections from anyone with verifiable information. We do not accept anonymous additions without source documentation. All corrections are reviewed before integration and credited where the source consents.