Pattern Recognition of Gold and Mercury Supply Chain in Global Trade Data

Despite the Minamata Convention’s targeted reductions in mercury consumption, global trade data exhibits a ‘Compliance Paradox’ where reported flows vanish while artisanal gold mining output remains stable. This research proposes a ‘Mineral Intelligence’ pipeline utilizing unsupervised machine learning to detect illicit mercury trafficking disguised as Electronic Waste (HS 8549). By applying Gaussian Mixture Models (GMM) and Isolation Forest algorithms to UN Comtrade data (2020–2024), we identify a systemic ‘Balloon Effect’: as elemental mercury bans took effect in 2022, illicit volumes were structurally displaced into ‘fake waste’ classifications. Forensic analysis reveals a statistically significant ‘Smuggler’s Signature’ within these flows, characterized by a price anomaly of $24–$80/kg (mirroring liquid mercury markets) and a Net-to-Gross weight ratio exceeding 90%, physically corresponding to standard 34.5 kg steel mercury flasks. Furthermore, Node2Vec and spectral embedding analysis exposes a ‘Decoupling Chasm’ (Manifold Distance: 2.06) that topologically separates financial gold hubs from mercury-intensive mining zones. Finally, Recursive LSTM forecasts predict a ‘burnout’ of the current HS 8549 smuggling vector (-618M kg/yr), warning of an imminent regime shift toward chemically masked commodities.