The global roots of pre-1900 legacy mercury

Significance

We present a comprehensive historical dataset on pre-1900 global trade and production of mercury, a country-by-country, chronologically exact, accounting of global mercury markets and estimated production losses, required for an accurate evaluation of the impact of pre-1900 legacy mercury on the global biogeochemical cycle and environmental inventory of mercury. A critical review of previous interpretations of historical data explains why they overestimated emissions of pre-1900 legacy mercury. The chemistry of silver refining with mercury and the more diversified market for mercury prior to 1900 requires differentiating legacy mercury from total historic anthropogenic mercury, to account for chemically sequestered mercury in end or by products, and the potential of delayed releases and emissions of mercury from those products.

Abstract

During the nineteenth century, a major change took place in the trade, production, and use of mercury that altered its nearly exclusive link to silver refining in the Hispanic New World. We track the global expansion of mercury markets in chronological detail from 1511 to 1900 using historical archives on production and trade, a detailed country-by-country accounting of the pool of anthropogenic mercury from which legacy mercury was ultimately generated. The nature and profile of pre-1900 legacy mercury extends beyond silver refining, mercury production, and gold extraction, and includes alternate sources (vermilion, felt, mercury fulminate) and new regions that were not major silver or gold producers (China, India, United Kingdom, France, among others), that accounted for approximately 50% of total mercury consumed in the nineteenth century. The nature of the pre-1900 mercury market requires a quantitative distinction between legacy mercury and historic anthropogenic mercury production and use, since the chemistry of its end-uses determines the pathways and timelines for its incorporation into the global biogeochemical cycle. We thus introduce the concept of a mercury source pool to account for total historic anthropogenic mercury within and outside this cycle. Together with a critical review of previous assumptions used to reconstruct the historical use and loss of mercury, a much lower level of emissions of pre-1900 legacy mercury is proposed. A coordinated effort across disciplines is needed, to complete a historically accurate scenario that can guide the multilateral policies adopted under the United Nations Minamata Convention to control mercury in the environment.

Mercury (Hg) has been transported across oceans and used in industrial quantities since the Hispanic New World was found to contain the world’s largest primary deposits of sulfidic silver ores. As early as the 1580s, the silver (Ag) refining ingenios in the vicinity of the city of Potosí in Upper Peru (now Bolivia) were consuming nearly 0.35 Gg of Hg per year (Gg/y), and total global production of Hg increased to an average of 4 Gg/y by the end of the nineteenth century (19C) (1, 2). By mid-20C the global community recognized the need to protect human health and the environment from anthropogenic Hg and its compounds, ultimately leading to the United Nations (UN) Minamata Convention on Mercury (2017) to curtail the mining and use of Hg worldwide (3).

In 1993, Nriagu brought attention to the major quantities of Hg employed to refine Ag and gold (Au) in the Americas prior to 1900 (4). Basing his estimate on global Hg production, he proposed a total loss to the environment of 126 Gg of Hg between 1580 and 1820 due solely to the refining of Ag with Hg, which included emissions to the atmosphere, and releases to soil or waterways during refining or in transport. For the period 1820 to 1900 he stated that 70% of reported silver production from Hispanic America required the use of Hg. He calculated the amount of Hg employed for that purpose by applying an empirical weight ratio (WR) of Hg consumed to Ag produced equal to unity. He thus estimated a total loss of 70 Gg from 1820 to 1900. To estimate the amount of Hg emissions, he applied to these total losses an emission factor (EF) of 60 to 65%, a range he justified by conflating historic silver refining with modern Artisanal and Small-Scale Gold Mining (ASGM) practice in the Amazon (4). In a second paper Nriagu added his scenario for the United States from 1850 to 1900, Table 1 (5).

Table 1.

Comparison between Hg SP data and main estimates published

Reference	Period	Total E+R+CS Hg	E	R	CS Hg	Location	Data and Assumptions
This paper	1501 to 1810	222	23 to 24	65 to 97	102 to 133	Americas, Europe	Values based on country-specific Hg trade and production data. From 1501 to 1810 all documented Hg production deemed used for Ag refining, of which 85% chemically sequestered as calomel, 12% as releases, 3% as emissions. 34 Gg of unaccounted-for Hg has been assumed either to have been lost in transit (maximizing estimated R) or 5% lost in transit and the remainder used for Ag refining.
	1811 to 1900	258	No breakdown possible in the absence of documentary archival data on Hg mass balance from end uses in 19C	Americas, China, Europe, India
	1501 to 1850	282
Nriagu (4, 5)	1570 to 1820	126	76 to 82	44 to 50	_	Hispanic New World	85% of Hg produced at Almadén plus all production from Idria and Huancavelica deemed used for Ag refining. A weight ratio of Hg loss to Ag produced (WR) of unity. 60 to 65% estimated as emissions and 35 to 40% releases (including losses in transit)
	1820 to 1900	70	42 to 46	25 to 28
	1570 to 1900	196	118 to 128	69 to 78
Nriagu (5)	1850 to 1900	61	37	24	_	USA	For USA, 90% of domestic Hg production plus imports deemed used to produce Ag and Au. 60% lost as emissions.
Pirrone et al. (6)	1800 to 1920	_	55	_	_	USA	As per Nriagu (4, 5)
Strode et al. (7)	1870 to 1880	_	_	_	_	USA	The sum total of Ag and Au production is multiplied by a common WR of 1.5, and by a common emission factor of 60%. Emissions based on these assumptions were then reduced by 50%.
Streets et al. (8)	up to 1850	_	137	_	_	Americas, Europe	Same WR of 1.3 is applied to Ag/Au production totals to estimate Hg use. A common pre-1850 emission factor of 52% is applied. Loss of Hg during production (PL) estimated as 18% of Hg produced; other sources included
Horowitz et al. (9)	1850 to 1900	Only in graph form	Not significant prior to 1900	Americas, Europe	Base assumption is that prior to 1900 almost all Hg production was used to produce Ag and Au. As per Streets et al. (8) for Hg used for Ag and Au production
Streets et al. (10)	up to 1850	416	136	280	_	Americas, Europe	Pre-1850 based on Amos (11), also Streets et al. (9)
Streets et al. (12)	up to 1850	362	53	277	Formation of calomel acknowledged	Americas, Europe	Streets et al. (12), maintain Nriagu’s assumptions for Hg used for Au production in USA in 19C (5)

For the assumptions used to estimate emissions (E), releases (R), and chemically sequestered Hg (CS Hg), see text and SI Appendix, Table S1.

The cumulative amount of Hg added to the global atmosphere from Ag and Au production in North America from 1800 to 1920 was then estimated by Pirrone et al. to have been about 60 Gg, versus some 10 Gg from industrial sources from 1920 to 1998 (6).

Later work continued to cite Nriagu’s initial set of assumptions, though Nriagu’s distinction that only 70% of pre-1900 Ag production required Hg was no longer mentioned. His conflation of historic Ag and modern ASGM processes using Hg remained unquestioned and reflected in calculations based on sum totals of deemed Ag and Au production, with a higher WR of 1.3 to 1.5, and an EF of around 60% applied to estimate joint Hg emissions (2, 4–12). Additional sources of pre-1900 Hg issued to the environment (such as Hg production, fossil fuel combustion, municipal waste incineration, smelters, vermilion/pigments, explosives, among others) were incorporated into global biogeochemical cycling models (hereafter referred to as Hg models). However, in all modeling scenarios the main contributors of pre-1900 losses of Hg continued to be Ag refining, Hg production and Au extraction. The common thread to the cited research effort was that all Hg produced in any given historical period ended up in the environment (2, 4–13). Notably, North America and Europe were predicated as the main contributors of pre-1900 anthropogenic emissions of Hg to the environment, while Asia was deemed to be only a minor contributor (10, 12).

Within the modeling work, a major pulse of Hg emissions from Ag and Au production in North America in the late 19C was proposed, coinciding with the gold rush of the period (8, 10, 12), but a majority of measurements of Hg in natural archives failed to detect it (6, 14–17). The absence of a peak in air deposition of Hg in a lake close to Potosí for the period of peak historic silver refining was reported as an unexpected finding (18). In parallel, studies on the correlation between the chemistry of the silver refining process based on Hg, and the historical ratio of mercury consumed to silver produced were published. It was known since the second half of the 19C that the critical role of Hg during the refining of silver ores was the reduction of silver halides (mainly chlorides) to elemental silver and thus the formation of calomel (Hg₂Cl₂) (19, 20). This was reconfirmed in a modern laboratory study of the complex chemistry of this refining process, identified as an early example of hydrometallurgical chloride leaching (21). A mathematical analysis of the relation between the nearly constant historical WR of Hg to Ag produced and the stoichiometry of the process concluded that emissions and releases of Hg represented only around 15% of total Hg used, due to the nature of the silver ore being processed (22, 23). Of this amount, a review of first-hand observations and measurements carried out in the 19C concluded that emissions reported under normal operations accounted for a very low percentage (estimated at 1 to 3%) of total Hg use (23). The remaining 85% of Hg used for refining silver ores was converted to solid calomel, washed away within the waste mineral matrix (23). These findings were incorporated in more recent models and reviews to explain lower levels of historical Hg air deposition measured in natural archives (12, 15, 24).

The challenges involved in modeling the biogeochemical cycling of Hg were reported: “Balancing the sources, sinks, and fluxes of Hg in different parts of the world at different times remains a modeling challenge, however, and lack of knowledge of historical man-made contributions has inhibited progress.” (8), while researchers faced “difficulties for the modeling of biogeochemical Hg cycling, because it was known that pre-1850 anthropogenic emissions were significant, but it was not known how large they were nor where they were located” (12).

In the following sections, we address this impasse by submitting a detailed set of historical data for the period 1501 to 1900 that provides a detailed quantitative picture of the global spread of diversified Hg end-users during this time frame. Prior to 1501, there are no historical records that provide reliable and continuous data on production and end-uses of Hg, which in any case are insignificant compared to later years. At the other end of our period of study, previous historical studies for the 19C had already identified the importance of London as a Hg reexport hub, with emphasis on pre-1850 global destinations, and of China for Hg exports from the United States after 1850 (25, 26). We complement these initial efforts by completing the data on the global trade and production of Hg to cover the whole 19C. The turn of the century in 1900 signals a major shift in the narrative of Hg losses to the environment, since cyanide substitutes Hg in Ag refining, and other sources of Hg in the environment now take precedence, such as power plants burning coal, emissions from smelters and other industrial sources, and ASGM.

Our 400-y span of detailed data on an increasingly diversified Hg global market lead to a substantive revision of the nature, location, and magnitude of historic Hg emissions and releases, include Hg by- and end-products in the accounting of the impact of historic uses of Hg on the environment, and provide guidelines for the interpretation of long-term Hg enrichment factors measured from natural archives.

Results

The historical data are presented using the terms defined as follows:

Emissions (E): emissions of Hg or Hg compounds to the atmosphere.

Releases (R): Hg issued to soil or water.

Chemically Sequestered Hg (CS Hg): mercury that is chemically incorporated in by-products of its use, such as calomel, or in manufactured end products such as vermilion and felt (see below). We ignore in this discussion waste from the smelting of cinnabar ores. The stability of these Hg compounds, and their use, will determine the timeline in which CS Hg is reinserted into the global biogeochemical cycle. Chemical sequestering can remove an important fraction of historic anthropogenic Hg from having any significant influence on modern background levels, such as for example vermilion. Our use of the term CS Hg thus differs from the use of the term “sequestered” in other work, where it is used to describe “unaccounted Hg … in landfills and other localized waste depositories” (8), or “some portion of the releases to land/water are undoubtedly immobilized close to the point of release, consistent with archival records and the large reservoirs of Hg observed to be stored at contaminated sites. Our best estimate is that approximately 40% of land/water releases are sequestered …” (10), a use of the term that does not include such an important manufactured product as vermilion, for example. We do not consider that Releases include CS Hg, since the latter encompass an unknown but not insignificant fraction of Hg that may be abstracted for long periods of time from participating in the global biogeochemical cycle of Hg.

It is important to point out that the extent of chemical sequestering of Hg can vary substantially across all pre-1900 end-uses of Hg. From the information currently available, it is greater in silver refining or the production of vermilion. For all end-uses, in addition to the possibility of Hg being chemically sequestered via a chemical reaction, elemental Hg was also released or emitted during production or from the end-product itself, to different extent according to each product and process. Also Hg that was initially chemically sequestered would swiftly revert to the environment upon use of the product (mercury fulminate on explosion). Fig. 1 schematizes the main paths of anthropogenic Hg from refining of cinnabar ore and its ultimate end-uses, both immediately to the environment as E and R, as well as delayed as CS Hg. Each end-use has a different balance of E, R, and CS Hg, as denoted in the width of the arrows in Fig. 1. Applications such as the production of mirrors or medical uses would have a higher contribution of E and R than CS Hg, while the large consumers of Hg such as silver refining and vermilion production would contribute higher levels of CS Hg than E or R.

Fig. 1.

The historic Hg SP encompasses Hg PL, and the Hg market, both datasets available from archival documentary sources, together with losses in transit or storage. Legacy Hg is less than or equal to Hg SP, subject to the fraction of Chemically Sequestered Hg (CS Hg) that remains stable over time.

Legacy Hg: Due to its longevity in the environment, all Hg produced in the past can potentially persist in the environment well after it was produced, and Legacy Hg is that fraction of historic anthropogenic Hg that actively participates in the global Hg biogeochemical cycle until it finally reposes in deep reservoirs (27). However, due to a significant fraction of total Hg production prior to 1900 destined for the manufacture of vermilion, or entombed as calomel during the refining of silver, any estimate of Legacy Hg must exclude CS Hg.

Production Losses (PL): emissions and releases of Hg during the processing of cinnabar ore (HgS) to extract Hg.

Hg Source Pool (Hg SP): For each country, we introduce the concept of a Hg SP, as the sum total of anthropogenic Hg in any given year, from where legacy Hg will be sourced. It is calculated from the historic data contained in documentary archives, by year:

Hg SP = Hg market + PL + unaccounted-for Hg

where

PL = Production Losses, estimated as a percentage of Hg production, and

Hg market = Domestic Production + Imports – Exports

Unaccounted-for Hg = Difference between production and trade data

Since the regional distribution and end-uses of unaccounted-for Hg are not documented, in order to present a detailed regional breakdown Hg SP* denotes a HgSP that excludes unaccounted-for Hg.

Thus, Legacy Hg is the amount of Hg SP that is not CS Hg and can impact the global biogeochemical cycle as emissions and releases:

Legacy Hg = Hg SP – CS Hg = E + R

The role of inventory stocks is ignored, as discussed further on. Fig. 1 visually summarizes the relation between these terms. By presenting the historical data in this manner, modelers can run sensitivities where Legacy Hg can span values up to 100% of Hg SP, until such time a quantitative accounting of CS Hg becomes available.

Vermilion is the term used for the red pigment with the same chemical formula as cinnabar, used to prepare vivid red paints and for lacquerware (28).

Felt for Hats: The production of felt from animal pelts other than beaver required the use of a solution of mercury nitrate (HgNO₃) to obtain the required compacting between fibers, a process known as carroting, employed since the 18C (29).

Mercury Fulminate was first synthesized in 1800 and found its major commercial application by Alfred Nobel to detonate explosives (30).

Mercury Global Production, Trade and Uses, 1501 to 1811.

From mid-16C to early 19C, the production of Hg in Almadén (Spain), Idria (modern Slovenia), and Huancavelica (Peru) was driven primarily by the demand from Ag refining in the Hispanic New World (Fig. 2 and SI Appendix, Table S1) (31–34). Up to 1810, the global Hg SP had increased by a factor of 18 from its level in the 1500s, in two stages of accentuated growth (Fig. 3). Overall, the Hg SP of the northern hemisphere (NH) for the period 1501 to 1811 was on average 1.7 times higher than that of the southern hemisphere (SH) (SI Appendix, Table S1). However, the main sources of legacy Hg were not global, but very specific to Ag-refining areas in the Hispanic New World, and in Hg production sites in Europe and Peru (Fig. 2).

Fig. 2.

Global Hg production, markets, estimated PL, and unaccounted-for Hg, in Gg, 1551 to 1810. For tabulated data and sources, see SI Appendix, Table S1.

Fig. 3.

Hg SP derived from main Hg markets, PL by hemisphere, and unaccounted production, in Gg, 1551 to 1810. For tabulated data, see SI Appendix, Table S1.

Mercury Global Production, Trade, and Uses 1811 to 1900.

In the 19C, major changes took place, with new Hg production sites and a truly global Hg market (Fig. 4). California mines (US) became leading producers from 1850, Huancavelica wanes as a source, and secondary producers appear (Italy, Russia, Mexico) (2, 33, 35–40). A more complex global market supersedes the previous one dominated by Ag refining, now including the production of vermilion, felt, mercury fulminate, medical products, mirrors, and scientific instruments. This expanded global market was fed from three main trading hubs: a) the United Kingdom for Hg sourced from Spain, Italy, and partly Idria, b) the United States for Hg sourced from California, and c) direct exports from Idria.

Fig. 4.

Global Hg production, markets, estimated PL, and unaccounted-for Hg, in Gg, 1811 to 1900. For tabulated data and sources, see SI Appendix, Tables S4–S11.

The UK export market and domestic consumption, 1811 to 1900: In this period, the UK gradually evolved to become a major export hub for Hg sourced mainly from Spain (25), and also developed into one of the most important end-users of Hg in Europe, though the exact breakdown of its internal consumption is not yet determined. SI Appendix, Table S2 covers its exports by destination and also an estimate of its domestic Hg market (41–45).

The US export market and domestic consumption, 1851 to 1900: The major role of the United States begins with the onset of Hg and Au production as of the 1850s. SI Appendix, Table S3 presents an overview of US exports from 1851 to 1900. Exports to Europe were minimal, except for the United Kingdom. Australasia received low quantities while Africa close to nil. Over the whole period, 54% of Hg produced and imported was destined for domestic uses, and 46% was exported, mainly to Mexico and China (82% of total exports) (26, 35, 46–48).

Domestic market and exports from Idria, 1811 to 1900: The case of the third major source of Hg, Idria, stands out due to the high percentage (35%) of its production used to produce vermilion at the site, from 1801 to 1879 (49). The remainder was exported to the United Kingdom and to unidentified destinations in Europe and Asia.

Mercury Source Pool, Mercury Market, and Production Losses (PL) by Country and Region, 1811 to 1900.

The main Hg markets and PL by countries and regions for the period 1811 to 1900 are set out in Fig. 5 (SI Appendix, Tables S4–S11). In Europe, the Hg market (33.7 Gg) is of a magnitude comparable to that of the United States, and comprised the production of vermilion, felt, mercury fulminate, and other products (including a minor fraction for Au extraction). The presence of important Hg production sites (Almadén, Idria, Monte Amiata) made Europe into an important source of legacy Hg, with an estimated PL (26.6 Gg) that makes up nearly half (44%) of its Hg SP* (60.3 Gg) over the whole period.

Fig. 5.

Hg market and production loss (PL) by selected countries, and HgSP excluding unaccounted-for Hg (Hg SP*) by hemisphere. For tabulated data and sources, see SI Appendix, Tables S4–S11.

The North American Hg market (71.1 Gg) is divided between Ag refining in Mexico (32.4 Gg) and multiple uses in the United States (38.7 Gg). Estimated PL (13.9 Gg) accounts for around 18% of its Hg SP*, a consequence of Hg production in California. The domestic market by end-user in this period is unknown, since the first quantitative breakdown for the use of Hg in the United States dates from 1917. There is one estimate that, up to the 1890s, 10,000 flasks were used for precious metal refining, 8,000 for vermilion production, 2,500 for fulminate and pharmaceuticals, and 1,000 in miscellaneous uses. In other words, by the last decades of the 19C, over half of Hg production in the United States was destined for uses other than Ag refining and Au extraction (50). There is a broad increase in Hg SP* from the 1870s to 1890s, but no singular pulse in its magnitude. Its value increases approximately two-fold during this period, and then falls back to 1850s level.

In South America, the Hg market (15.1 Gg) is due mainly to Ag refining activity in Peru/Bolivia (7.3 Gg) and Chile (4.8 Gg). Estimated PL (0.6 Gg) accounts for just 4% of its Hg SP*, a reflection of the significant decrease in Hg production in Huancavelica. This region was not a major source of Hg emissions for this period.

In Asia, the overall Hg SP* (47.6 Gg) is dominated by China’s Hg market (31.6 Gg), on par with Mexico and just under the United States. Production of vermilion would have been an important end-use for Hg in China, but we have no information on the breakdown of its internal Hg market. India has a sizeable Hg market (6.9 Gg), though we could not determine its nature. Estimated PL for Asia (2.2 Gg) is a minor contributor (5%) to its total Hg SP* (47.6 Gg), based on current estimates of China’s production of Hg (2).

The total Hg SP* in NH (236 Gg) dominates its counterpart from SH (19 Gg), as well as the estimated PL of the NH (43 Gg) with respect to the PL for SH (0.6 Gg). In contrast to the 1501 to 1811 period, the second half of the 19C is a period of global dispersion of potential sources of legacy Hg (compare Figs. 2 and 4). The split in global Hg end markets is evenly matched between countries with major Ag/Au production and those without.

The irruption of China as a major consumer in the global Hg market is highlighted in Fig. 6, increasing its use as the United States decreased theirs toward the end of the 19C. Finally, neither Australasia (2.6 Gg) nor South Africa (0.6 Gg) are major contributors to the global Hg SP* up to 1900. Though the former was a major Au producer in this period, on par with the United States, we explain below why its use of mercury was limited in the early stages of its Gold Rush. From 1850 to 1900, the global Hg SP* increases approximately fivefold (SI Appendix, Table S11).

Fig. 6.

The HgSP excluding unaccounted-for Hg (Hg SP*) of selected countries, from 1811 to 1900, from where legacy Hg was generated. For tabulated data and sources, see SI Appendix, Tables S4–S11.

Discussion

The market Hg dataset is based on real-time trade accounts for which there is no similar set in scope in the published literature. The Hg production data from 1510 to 1900 (SI Appendix, Table S1 and S4) are within 2% of the total published by Hylander and Meili (2). Our estimates of PL are based on documentary evidence of the period (see below). In the absence of sufficient information on the evolution of historic Hg emissions with technology or chemical composition of ores, from smelters of nonferrous metals, waste incinerators or fossil fuels, we have not included them, to avoid anachronistic projections of EF. These are important sources of Hg losses post-1900 but not during the periods we have covered (12).

For the period 1501 to 1810, reported global production data reached 158 Gg while global trade data from the sources consulted were 120 Gg (SI Appendix, Table S1). From the former, we estimate a total PL of 63 Gg, which together with 34 Gg of unaccounted-for Hg indicate a minimum value of Hg SP of approximately 183 Gg. In this period, Ag refining by Hg and Hg production are the dominant sources of legacy Hg. Of the amount of Hg used for silver refining, as explained above, approximately 85% of Hg was issued to the environment in the form of solid calomel, and less than 3% as emissions of volatile Hg. Around 12% would have been released as liquid Hg spills to the ground or entrained to waterways (23). The chemical evolution of calomel in the soil over time remains to be established, and the conditions under which it could disproportionate into elemental Hg and mercury sulfide (HgS) are currently under study at historic mining sites (51). As to the breakdown of PL, until more historical data become available (see below), we can only estimate emissions from Hg production at 30% and releases (losses to waterways and soil) at 70%.

To arrive at a range of Hg emissions, releases, and CS Hg for this period, we need to make assumptions regarding the fate of the unaccounted-for 34 Gg, the difference between documented production and trade data. We chose one scenario whereby all 34 Gg were lost in transit, thus maximizing total releases. The second scenario assumes 5% of unaccounted-for Hg was lost in transit, and the remainder was employed to refine Ag, thus maximizing CS Hg. As observed in Table 1, for the period 1561 to 1810 the change in scenarios barely impacts estimated emissions (23 to 24 Gg), contrary to its impact on the estimate of releases (64 to 96 Gg) and CS Hg (102 to 133 Gg) (see also SI Appendix, Table S1).

The approach by other research groups has been to assume that all Hg in this period has been destined for Ag production (which is a historically valid approximation) and that an EF around 60% can be applied to these data to estimate emissions (which is not, since the chemistry and technology of the process indicate a more probable value of 3%, as discussed above), or else total Ag production has been multiplied by a deemed WR (1 to 1.3) to estimate Hg consumption, to which the same EF range is applied to estimate emissions (4, 5, 7–10, 12, 13). This methodology compounds two major errors. Though Nriagu made the historically correct observation that only a part of total Ag production was based on the use of Hg (4), subsequent authors (7–13) used data on total Ag production from sources that do not identify how much Ag was produced via smelting with lead (Pb) and how much via refining with Hg (52). Using the sum total of the two to calculate use of Hg is not a minor error. Prior to 1811, at least 30% of Ag produced in New Spain did not require Hg (23), and in Europe and Japan Hg played an insignificant role in the production of Ag (53). Furthermore, though the fate of calomel is still uncertain and its disproportionation in time cannot be ruled out (12, 23, 51), the same EF cannot be applied to joint Ag and Au production data, since the former involves a complex chemical process, while the latter is a simple amalgamation. For Ag refining EF is approximately 3%, as discussed above. For gold extraction, Nriagu cites an EF of 65 to 83%, as evidenced in late 20C ASGM in the Amazon (4). This estimated EF for Au cannot be conflated with that of Ag, since it is an anachronism to correlate a range observed in the 20C for gold extraction to a completely unrelated silver refining process carried out from the 16C to 19C.

For the period 1811 to 1900, reported global production data reached 215 Gg while global trade data from the sources consulted were 169 Gg (SI Appendix, Tables S4 and S11). From the former, we estimate a total PL of 43 Gg, and together with 46 Gg of unaccounted-for Hg we arrive at a minimum value of Hg SP of 258 Gg for this period. It is not possible to propose a breakdown into E, R, and CS Hg, because there are not sufficient historical documentary data on Hg end-uses in this period. Since approximately half of the global Hg market was in countries with no or minor Ag/Au production, this gap in knowledge is substantial and questions any reconstructions proposed in previous modeling for this period. In addition, published estimates have misinterpreted historic Ag production and Hg consumption data for the latter part of the 19C in the United States, thus contributing on three fronts to an overestimation of Hg emissions. First, the main metallurgical process to refine Ag after 1875 was smelting, not the use of Hg (Fig. 7) (54, 55), so previous projections of Hg consumption based on total Ag production data of the United States (5, 10, 12) have overestimated its use for that purpose. Second, the assumption that 90% of domestic production plus imports in the United States was used for Ag and Au production in the United States (5, 12) is not correct, since nearly half of production was exported. Third, the chemical sequestering of Hg in end products such as vermilion, or in by-products such as calomel have also been omitted (8–10, 12), which has led to higher estimates of emissions and releases.

Fig. 7.

Fraction of Ag produced in the United States by smelting, 1875 to 1900. Derived from data from ref. (54) and ref. (55).

The deemed impact of gold production on emissions of pre-1900 legacy Hg (8, 10, 12) also needs to be reexamined based on our trade data. The United States and Australasia from 1850 to 1900 produced similar quantities of Au (56, 57). Since Australia had no significant domestic Hg sources (58), nor any major end-uses other than Au extraction, the amount of Hg imports is a very good measure of the consumption of Hg for its production of Au. As clearly evidenced in Fig. 8, during the initial peak of the Australasian Gold Rush, between 1851 and 1858, there was very limited imported Hg. Human skill and the nature of the deposits combined to generate major production output without much need for Hg. In total, during the second half of the 19C, 2.6 Gg of Hg was imported to produce just under 3.3 Gg of Au (SI Appendix, Table S11, 56, 57). If this ratio is extrapolated to the production of Au in the United States, it would imply that less than 10% of the Hg market in the United States was consumed for this purpose. The fact that Au was also produced from the refining of Ag would further lower this percentage (19). Data from Australia cannot be extrapolated directly to the United States since the nature of Au deposits, abundant local Hg, water resources, and techniques also determined Hg consumption, but it certainly cautions against identifying the California Gold Rush as a major source of emissions of legacy Hg in the latter part of the 19C. It also agrees with new assessments of lake sediment core data (14, 17) and ice core results (59) from locations in the western United States, which do not show a strong increase in Hg emissions and subsequent deposition to western US watersheds during the latter part of the 19C.

Fig. 8.

Gold production in Australasia vs Hg imports, 1851 to 1860 (56, 57) and SI Appendix, Table S11.

When we compare previous projections with our data (Table 1), the significant overestimation of pre-1900 Hg emissions for the reasons explained above is made clear. Even though the different periods in time do not match exactly, proposed levels of Hg emissions range from twice to five times our own calculation. The proposed global yearly peak of (E+R) in 1890 of around 8.2 Gg (10, 12) is double the global Hg SP* value for the years 1888 to 1892, which is not realistic. The most recent estimates of total Hg loss (E+R) up to 1850 (10, 12) exceed by nearly 30 to 50% the global Hg SP for that period.

Overall Table 1 questions the assumption of a large contribution to the global inventory of Hg from pre-1900 emissions of Hg, and evidence the omission of important amounts of CS Hg from previous models. More research is required to determine which of the varied end-uses for Hg in the late 19C could have produced major injections of Hg directly into the atmosphere from thermal processes, other than the minor losses from Ag refining. With our current state of knowledge only emissions from Hg production sites are deemed the major source of emissions of pre-1900 legacy Hg. Releases to soil and waterways, and recondensed emissions in the vicinity of Hg production centers or end-users, remain a more credible historical scenario for the majority of pre-1900 legacy Hg.

Additional Guidelines Provided by Hg SP Datasets.

The geographical and chronological detail of our dataset constitutes an independent set of values to cross-check results offered by models. Thus pre-1900 Hg total yearly emission and releases by country produced by models cannot exceed each country’s yearly Hg SP* values, since this would imply the total conversion of Hg consumption and losses into emissions, and the omission of any CS Hg. The global limit is 2.5 Gg/y of total losses (SI Appendix, Tables S1–S11). There is also no evidence of a major pulse in the Hg SP of the United States around the 1880s that would be lost completely as emissions (Figs. 5 and 6) (8–10, 12).

On a regional level, our Hg market and PL data evidence a global spread of legacy Hg sources (Fig. 4) that has not figured in reconstructions from modeling studies, which only propose the Americas and Europe as main contributors to legacy Hg prior to 1900 (8–10, 12). China and India were important consumers of Hg that need to be brought into the discussion, while Europe was not only a major producer but also a major consumer of Hg for uses other than Ag or Au production.

The enrichment factor for Hg in the environment has been calculated from measurements of Hg flux and concentration in natural archives over different periods (17). Our data offer a different option, the possibility to measure the increase and geographical spread of Hg SP* spanning four centuries, with a chronological detail and exactitude unattainable in samples from natural archives. We identify at least two significant step increases in Hg SP prior to 1850 (Fig. 4), leading to an increase of around 18 times in Hg SP from 1501 to 1900. There is no single pre-1850 Hg SP base line. The absence of sufficiently temporally resolved records from natural archives make it impossible at this time to determine which periods of increased Hg SP along this 400-y span can be considered to be still relevant to the discussion on legacy Hg.

Most estimates of Hg enrichment factors from natural archives have used 1850 as a reference point (14, 17), yet the period around 1850 poses a challenge since it encompasses a sharp increase in Hg SP* values. For example, if we take as a reference point the values of Hg SP* in the 1850s, in the United States, there is a twofold increase by the 1880s, which falls back to around unity by 1900 (Fig. 5 and SI Appendix, Table S12). If on the other hand, we use as reference the average for the previous decade, 1840 to 1850, the average Hg SP* for 1890 to 1900 increased 25 times (SI Appendix, Table S12). The correlation of these values with reported ratios of Hg flux measured in estuary or sediment cores from remote lakes (2× or less in most locations, up to 5× in lakes of the Sierra Nevada, derived from published plots), is either excellent or nonexistent, subject to the level of uncertainty in the chronology of natural archives (14, 17, 60). The range of these measurements in countries with high pre-1900 Hg SP* values (Mexico, China, India, U.K., France, among others) is too limited at present to arrive at conclusions on the correlation between Hg SP and local Hg enrichment factors over time.

Conclusion

The use of Hg from 1500 to 1900 evolved from a near monopoly by colonial silver refiners in the New World to a global market in the 19C that included China and India as major consumers, together with a wider range of end-uses, from silver and gold production to vermilion and felt hats. This introduces a fundamental change in the use of historical data in models that study the impact of pre-1900 legacy Hg on the environment. A significant amount of anthropogenic Hg prior to 1900 was removed from participating in the global Hg biogeochemical cycle since it was chemically sequestered either in stable manufactured products such as vermilion or as a by-product such as calomel buried within a mineral matrix. At present, we do not have enough historical information to prepare quantitative substance flow analysis diagrams, such as proposed by Horowitz et al. (9), for each manufacturing process that used Hg in the 19C. We therefore introduce the concept of a Hg SP as a way to profit from the detailed global trade and production data available from documentary archives that allows boundaries to be established to the magnitude to legacy Hg, based on country-specific and real-time historical archives. The degree of chronological and regional detail presented has not been available to models up to now.

In addition, our results indicate that the significant geographical spread of the Hg market in the 19C, to China, India, U.K., France, and others, that accounted for approximately half the global value of Hg SP in the 19C, needs to be included in future models. The rise of China as major consumer of Hg and producer of vermilion in the 19C, accounted for nearly 20% of the documented global Hg market during this period. This fact by itself, regardless of similar or other end-uses for Hg in the United States or Europe, means that a significant amount of Hg in the 19C was chemically sequestered as vermilion, and thus would not be part of the global Hg biogeochemical cycle.

The low consumption of Hg evidenced in Australian gold production is a strong argument against the proposal that the late 19C gold rushes were a major contributor of legacy Hg. Together with the fact that CS Hg played a major role in the mass balance of pre-1900 anthropogenic Hg, it explains the absence of supporting evidence from natural archives for major pulse of Hg emissions in the late 19C. Previous estimates of emissions from Ag or Au production have been overestimated, due to the omission of CS Hg, in the form of calomel, the conflation of data from two quite distinct processes (Ag and Au production based on Hg), or that the major exports from California to China have not been taken into account in previous models.

A more geographical diverse global effort is required to be able to correlate the data on Hg from historical and natural archives. The environmental impact of legacy Hg from hat making in Connecticut (US) has already been recognized (61, 62), but further work is needed for all alternate Hg industrial hotspots not related to Ag/Au/Hg production, such as Danbury and Norwalk (US), Paris (France), London (U.K.) for felt production, New York (US), Amsterdam (Holland), and unspecified locations in China for vermilion, among others.

A better documented historical estimate of losses (PL) at each of the main Hg production sites is also needed, since they are the main candidate as the major source of emissions of pre-1900 legacy Hg, to complement data from natural archives such as already reported for Huancavelica (63).

On the chemical side, since Hg consumed in Ag production was the main contributor to the global Hg SP from mid-16C to 1900, the life cycle of calomel in the soil must be established to determine the possible long-term chemical sequestering of Hg, and the end-products of any disproportionation (12, 51).

A coordinated effort across disciplines is needed to converge the modeling of pre-1900 legacy Hg with the historical and chemical reality of the Hg global market and production presented in this paper. This in turn will improve the accuracy of predictions on the impact of legacy Hg on the current environmental inventory of Hg and its biogeochemical cycle.

Methods

Mercury Market.

We considered the Hg market for each country as equivalent to the net import/export balance, together with the fraction of domestic production that was not exported (where applicable). The Hg SP was calculated as the sum of Hg market data plus PL (where applicable). Export destinations were registered by first port of call, so onward secondary distribution though unknown cannot be discounted but its impact on the conclusions is considered minimal (25). For the sake of brevity, destinations are identified by modern country names in all tables and figures.

Mercury Production Loss.

The only Hg production site that measured PL during operations was Idria. They reported an average PL (calculated as a percentage of Hg produced) from 1786 to 1879 of 19% ± 11%, which reflects the wide range of values in the available dataset (49). For other Hg production sites, there are no equivalent data sources, only guesses or very limited trials from contemporary observers. For all Hg production prior to 1811, therefore, a PL of 40% is applied, based on the persistent contemporary reports of deemed high losses (64–68). After 1811 a PL of 20% is adopted for every location, to reflect the wider use of Idria-style condensers, except for Huancavelica (Peru) which is kept at 40%, based on its older technology. PL comprised impregnation by Hg of the internal surfaces of condensers, unextracted Hg in soot, liquid Hg spills to the soil and Hg releases to the air (69, 70). Though each type of PL was not measured, Hg emissions would have been an important fraction, albeit at low heights. Since furnace design, condensers, and operational practice varied substantially between locations and time periods, the estimates used in this exercise are an initial approximation to what undoubtedly were significant losses during production.

Integrity, Quality, and Reliability of the Documentary Data Sourced for This Work.

Historical data have to be interpreted taking into account what Kahneman et al. have described as objective ignorance, the reality of an intractable uncertainty (for example, in our case contraband, which cannot be quantified with any certainty) and imperfect information (what in principle can be researched but is not yet available, such as historical production of Hg and vermilion in China, or details on the mass balance of Hg during production of vermilion, felt hats, and others prior to 1900) (71). Thus, our data on global production and trade summarized in Table 1 and tables in SI Appendix represent a base line, minimum values from primary (documents created in each period of interest, such as government reports and newspapers) and secondary (articles or books) sources. It is the lower limit because all pertinent but unquantifiable historical factors (contraband, omissions in shipping manifests, missing records) would add to the numbers reported in primary sources. We do not have sufficient historical data to establish un upper limit with any confidence. The range in our pre-1811 estimates in Table 1 is not indicative of error in the source data but illustrates how the interpretation of the unaccounted-for difference between production and market data has virtually no impact on estimated levels of emissions.

For data prior to 1811, a period amply covered in historical sources, the difference in magnitude between available official Hg production data and Hg markets in the Hispanic New World leaves approximately 25% of the former unaccounted for. There are various explanations possible. Historical data have been sourced from different registers, each with its own degree of accuracy, and does not consider trade in contraband, which is impossible to quantify, and losses in transit. In addition, we have no data on i) Hg exported from Idria to Europe or Asia ii) Hg used for Au extraction in Africa, Brazil, or Colombia iii) production of vermilion in Spain and Idria iv) from the 18C onward, use of Hg for felt in the United Kingdom and France, and v) inventory levels, which are deemed very low in this period (72–74).

For the global Hg trade between 1811 and 1900, we turned to the primary records kept by the UK government, both the very detailed customs records kept at the National Archives and accessible to the public via internet, cross-checked with those reported under the UK Parliamentary Papers for the period (41–45). In the case of other European data, approximately 31 Gg of production from Almadén (12 Gg) and Idria (19 Gg) remain unaccounted for, from a total European Hg production of 133 Gg. Up to 5 Gg of inventory has been reported during this period, though detailed information on inventory levels is scarce (35, 50).

In the case of the United States, primary sources include government publications, local trade associations, and also export cargoes by ship reported in newspapers (35, 46–48). The data for the first two decades (1850 to 1870) do not correlate well between available sources probably due to the incipient nature of record-keeping by the government bodies of the United States during this period. Thus, we estimate that up to 20% of the value assigned to the domestic market from 1851 to 1870 could also correspond to unreported exports. In total, we have tracked a global trade of 181 Gg of Hg between 1811 and 1900, compared to a total global Hg production for the period of 215 Gg.

The integrity and quality of the data employed stems from either the government imprimatur of the primary sources used, or the academic integrity of the secondary sources employed. We interpret reliability in the context of this study as consistency between independent documentary sources, global trade data on one hand, and production data on the other. The Hg trade data presented between 1501 and 1811 account for 75%, and for the 1811 to 1900 period 78%, of the corresponding Hg production data. We thus consider our historical trade data to be reliable and comprehensive within the limits of the discipline.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.