About This Project
An open-source intelligence dataset and analysis site tracking Iranian military operations against Israel and US/coalition targets.
Why this project exists, how AI is used to synthesise fast-moving OSINT data, and who it's dedicated to.
Project Overview
This is an independent, open-source OSINT (Open Source Intelligence) project that collects, structures, and visualizes data on Iranian missile and drone attack salvos against Israel and US/coalition targets. The project covers four operations in Iran's "True Promise" series — the first direct Iranian military strikes on Israeli territory in the history of the conflict.
The dataset tracks 45+ attack salvos across four rounds (April 2024 — ongoing), with 76+ data fields per salvo covering timing, weapons systems, targets, interception, casualties, and escalation dynamics. Data is structured in JSON with a formal schema, and also available as a queryable Neo4j graph database.
Rounds Covered
- True Promise 1 (Apr 13–14, 2024) — First direct Iranian attack on Israel. ~320 munitions, ~99% interception rate.
- True Promise 2 (Oct 1, 2024) — Ballistic-missile-only strike. ~200 munitions. First Fattah-1 hypersonic combat use.
- True Promise 3 (Jun 13–24, 2025) — "Twelve-Day War." 22 salvos, ~1,700 munitions. 33 killed, 3,238 wounded in Israel.
- True Promise 4 (Feb 28 – ongoing, 2026) — Multi-theater expansion. 19+ salvos targeting Israel and US/coalition bases across 12 countries.
Data Sources
Data is cross-referenced from open-source reporting, official military statements, satellite imagery analysis, and verified social media:
- Wikipedia conflict timelines
- IDF, CENTCOM, and IRGC official statements
- FPRI, IISS, and defense research publications
- Times of Israel, Reuters, AP, Al Jazeera
- Satellite imagery analysis (Planet Labs, Maxar)
- AI-assisted multi-model LLM research with source grounding
Data Provenance
Different categories of data in the dataset originate from different source types. This section documents the provenance of specific data fields to support verification and reproducibility.
Launch Site Geolocation
Geographic coordinates for IRGC missile launch sites are derived from open-source intelligence on known Iranian ballistic missile bases. The primary source is the Alma Research and Education Center's January 2026 damage assessment, which documents 25 major MRBM launch bases across Iran with satellite-verified coordinates. Additional geolocation data is cross-referenced against:
- CSIS Missile Threat Project — missile system profiles and base locations
- NTI (Nuclear Threat Initiative) — facility profiles with coordinates (e.g. Tabriz)
- Wikipedia: Iranian underground missile bases — aggregated public domain data
Where the exact launch site for a specific salvo is unknown, coordinates are assigned to the most
operationally likely base in the western Iran corridor (Kermanshah, Khorramabad, Kangavar, Dezful,
Panj Pelleh) based on the weapon type, actor, and target. These are marked with
generic_location: true in the dataset. Launch coordinates are truncated to 2 decimal
places on the map for operational security reasons.
Target Geolocation
Israeli military installation coordinates are drawn from publicly available sources and truncated to 2 decimal places. US/coalition base locations are sourced from public Department of Defense disclosures and open-source mapping. Civilian impact locations are geolocated from media reports and are approximate.
Weapon System Identification
Specific missile and drone variants used in each salvo are identified from IRGC statements, IDF post-strike assessments, verified munition debris imagery (via OSMP), and defense analyst reporting. Technical specifications are drawn from CSIS Missile Threat, IISS publications, and manufacturer disclosures.
Interception & Casualty Data
Interception rates and system attributions are derived from IDF, CENTCOM, and coalition partner statements. Casualty figures are sourced from Israeli emergency services (MDA, Zaka), hospital reports, and verified media. Figures may lag or be revised as more information becomes available.
AI-Assisted Research
This project uses AI-assisted research (multi-model LLM workflows with source grounding) for initial data collection and structuring. All AI-generated data is cross-referenced against primary sources before publication. The site content itself (analysis, descriptions) is AI-generated and clearly marked as such.
Repository & Data Access
The source data is maintained in a public GitHub repository. The site you're viewing pulls data from the repo via an automated pipeline.
- Data Repository: github.com/danielrosehill/Iran-Israel-War-2026-Data
- Data formats: JSON, Neo4j (graph), CSV, GeoJSON, KML
- Schema: Validated against
data/schema/wave.schema.json - Data dictionary: Full field reference in
docs/data-dictionary.md - Kaggle Dataset: kaggle.com/datasets/danielrosehill/iran-israel-war-2026
- Hugging Face Dataset: huggingface.co/datasets/danielrosehill/Iran-Israel-War-2026
Contributing
Corrections, additions, and source citations are welcome via pull request or issue on the data repository.