Mercury and CO₂ emissions from artisanal gold mining in Brazilian Amazon rainforest

The study addresses how much mercury is currently used and lost in artisanal and small-scale gold mining (ASGM) in Brazil’s Tapajós River basin when retorts are employed, and quantifies the sector’s energy use and associated climate impacts. ASGM in the Amazon occurs largely in remote areas with limited law enforcement and complex, often inadequate regulation, leading to informal and illegal operations. Although the Minamata Convention requires National Action Plans to reduce mercury in ASGM, implementation in Brazil has lagged. Mechanization (e.g., excavators) is increasing, potentially changing energy intensity and emissions. The Tapajós region, the world’s largest ASGM district, features multiple mining methods (secondary deposits on land and by dredge; primary deposits underground and open pit), offering a setting to gather empirical, in situ data from typically inaccessible sites to evaluate mercury use/recovery and carbon footprints across techniques.

Prior research has extensively documented ASGM-related deforestation, social conditions, and mercury contamination of soils, waters, fish, and humans in the Brazilian Amazon. Policies governing mercury and ASGM in Brazil are complex, inconsistently enforced, and have changed over time, with earlier strict requirements on mercury recovery rescinded in 2015. Training programs since the 1990s promoted retort use, yet data on actual mercury use, losses, and retort effectiveness remain scarce and often based on small samples with comparability issues. Even fewer studies have addressed ASGM energy intensity and climate impacts; existing work largely examines individual mines outside Brazil with methodological inconsistencies. There is thus a clear knowledge gap on current mercury use and recovery rates in the Tapajós and on the carbon footprint of ASGM under different mining techniques and increasing mechanization.

Primary data were collected mainly during 2018–2022 across ~50 mining sites in the Tapajós region, including informal and illegal operations, using measurements, interviews, and questionnaires. Access was facilitated via known contacts and snowball sampling. Mercury use, loss, and recovery were quantified through 47 in situ mass balances following UNEP/PlanetGOLD guidance. To ensure comparability, fine gold (Au) content was characterized: sponge gold averaged 88% Au (from 12 sponge–doré paired datasets), while doré fineness was ~91% Au for secondary deposits on land (baixão), ~93% for dredges (dragas), and ~78% for primary deposits (filão), based on interviews with gold shop owners across six towns. Retort adoption was assessed via an anonymous survey (42 interviews: 27 garimpeiros, 14 managers/owners) in local establishments. Energy use was surveyed at 34 sites covering different mining types; fuel and electricity were converted to diesel equivalents using heating values. For the LSGM site Serabi, electricity was converted using site-specific diesel consumption for power generation (0.28 l kWh⁻¹). Logistics fuel demand (primarily for transporting fuel to garimpos) was estimated using satellite/georeferenced distances and interview data for vehicles and aircraft; average logistics was ~330 kg diesel kg⁻¹ Au (not included in Fig. 4 point values except Serabi). Explosives consumption for primary underground operations (~20 kg explosives kg⁻¹ Au) was included in impact calculations. Climate change impacts were estimated from measured/estimated fuel, logistics, and explosives using ecoinvent v3.9.1. Data quality challenges from non-standard bookkeeping were addressed by cross-checking with literature, machine datasheets, mass balances, and removing verifiably incorrect values; plausible outliers were retained.

- Mercury use and loss: Mean mercury use was 1.7 kg Hg per kg fine gold (Au). Mean mercury loss was ~0.19 kg Hg kg⁻¹ Au when retorts were used, reflecting losses via spillage during mixing/panning, leakage during retorting, and residual Hg in sponge gold. Retorts retained at least ~75% of Hg depending on retort quality and operator experience. In 88% of surveyed garimpos, retorts were in use. Some dredge measurements showed negative net loss (apparent recovery >100%), indicating dredges can collect legacy metallic mercury from sediments. - Regional emissions: Applying the 0.19 kg Hg kg⁻¹ Au loss to reported ASGM output in Pará (10 t in 2019; 18 t in 2020) yields ≥5 t Hg emissions across 2019–2020. Extrapolating to Brazil’s total 54 t gold production implies ≥10 t Hg emitted over 2019–2020 (optimistic, assuming universal retort use). Within the Tapajós, annual release was estimated at ≈2.5 t Hg even with retorts. - Energy intensity: Average diesel-equivalent energy intensities (kg diesel-eq kg⁻¹ Au) by mining type: primary underground (UG) 2,230; secondary land (overground, OG) with excavators 4,410; secondary dredges 6,410; secondary land OG without excavators 7,340. The nearby LSGM Serabi site was 3,820 kg diesel-eq kg⁻¹ Au and partly grid-powered (~40%). - Climate impacts: Corresponding carbon footprints (kg CO₂eq kg⁻¹ Au): primary UG 9,750; secondary land with excavators 18,000; secondary dredges 25,700; secondary land without excavators 29,200. Overall, ASGM lies roughly in the 10–30 t CO₂eq kg⁻¹ Au range, comparable to recent LSGM estimates (~21,000 kg CO₂eq kg⁻¹ Au from SKARN). Logistics contributed a relatively small share (~1,260 kg CO₂eq kg⁻¹ Au). Industrial recycling of gold had far lower climate impact (53 kg CO₂eq kg⁻¹ Au). - Mechanization trend: Increased mechanization (e.g., excavators) reduced specific energy/climate intensity via efficiency gains, although it may induce rebound effects (more sites opened, infrastructure expansion).

The findings demonstrate that retorts, widely adopted in the Tapajós, substantially reduce mercury losses compared with historical practices and represent successful technology transfer efforts since the 1990s. However, current recovery rates fall short of the former Brazilian regulatory target (96% Hg recovery), leaving environmentally significant emissions even with retort use. Observations of apparent Hg recovery >100% during dredging underscore the persistence of legacy metallic mercury in sediments from decades of ASGM; continued operations will likely remobilize this Hg regardless of current best practices, highlighting the need to better understand sediment–mercury dynamics and methylation conditions. Energy and climate analyses reveal that ASGM’s carbon footprint is high and comparable to LSGM, with mechanization lowering specific intensities but posing system-level rebound risks through accelerated expansion and infrastructure development. Certification and “responsible ASGM” initiatives focused mainly on mercury should incorporate greenhouse gas performance, and mitigation will require shifting away from fossil fuels toward renewables where feasible. Policy implications include the urgency for Brazil to complete and implement its Minamata National Action Plan, streamline formalization and permitting, and strengthen monitoring to reduce both mercury and climate impacts.

This study provides rare, in situ, multi-method evidence on mercury use/loss and energy/climate impacts across diverse ASGM methods in the Tapajós. Key contributions include quantifying average Hg use (1.7 kg kg⁻¹ Au), residual Hg losses with retorts (~0.19 kg kg⁻¹ Au), high retort adoption (88%), and substantial energy and CO₂ intensities that vary with mining method but remain high overall and comparable to LSGM. Policymakers should go beyond promoting retorts to ensure higher Hg recovery, adopt Brazil’s Minamata NAP, and formalize and monitor ASGM more effectively. Given that fossil energy dominates climate impacts, investment in renewable energy solutions is needed for both artisanal and industrial operations. Future research should prioritize large-scale surveys of sediment and soil mercury contamination, mechanisms of mercury methylation and remobilization, and broader system-level assessments to capture rebound and infrastructure effects of mechanization.

- Geographic scope: Results are specific to the Tapajós region and may not generalize to all ASGM contexts. - Measurement approach: Mercury assessments used mass balances without elemental composition, limiting attribution of losses to specific process steps and preventing direct environmental fate tracking; complementary soil/water/flora sampling and tools (e.g., handheld XRF) would improve resolution. - Data quality: Non-standard bookkeeping and literacy barriers required data validation and exclusion of incorrect values; plausible outliers were retained. - Retort usage variability: Retort adoption may be lower in certain illegal/indigenous areas than observed (88%). - Uncertainties in comparative benchmarks: LSGM climate data vary by technology, geography, and energy mix; estimates carry uncertainty. - System boundary: Transport logistics were estimated and not included in the primary energy intensity figure values (except Serabi), though accounted for in separate logistics estimates and climate calculations.