lable at ScienceDirect

Atmospheric Environment 43 (2009) 2012–2017

Contents lists avai

Atmospheric Environment

journal homepage: www.elsevier .com/locate/a tmosenv

Impact of mercury emissions from historic gold and silver mining:Global modeling

Sarah Strode a, Lyatt Jaegle a,*, Noelle E. Selin b

a Department of Atmospheric Sciences, University of Washington, Box 351640, Seattle, WA 98195, USAb Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA, USA

a r t i c l e i n f o

Article history:Received 17 July 2008Received in revised form7 January 2009Accepted 8 January 2009

Keywords:MercuryMiningGold rushNorth AmericaDepositionSediment cores

* Corresponding author. Tel.: þ1 206 685 2679; faxE-mail address: jaegle@atmos.washington.edu (L.

1352-2310/$ – see front matter � 2009 Elsevier Ltd.doi:10.1016/j.atmosenv.2009.01.006

a b s t r a c t

We compare a global model of mercury to sediment core records to constrain mercury emissions fromthe 19th century North American gold and silver mining. We use information on gold and silverproduction, the ratio of mercury lost to precious metal produced, and the fraction of mercury lost to theatmosphere to calculate an a priory mining inventory for the 1870s, when the historical gold rush was atits highest. The resulting global mining emissions are 1630 Mg yr�1, consistent with previously publishedstudies. Using this a priori estimate, we find that our 1880 simulation over-predicts the mercurydeposition enhancements archived in lake sediment records. Reducing the mining emissions to820 Mg yr�1 improves agreement with observations, and leads to a 30% enhancement in global depo-sition in 1880 compared to the pre-industrial period. For North America, where 83% of the miningemissions are located, deposition increases by 60%. While our lower emissions of atmospheric mercuryleads to a smaller impact of the North American gold rush on global mercury deposition than previouslyestimated, it also implies that a larger fraction of the mercury used in extracting precious metals couldhave been directly lost to local soils and watersheds.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Gold and silver mining in North, Central, and South America,and in Australia, which used mercury amalgamation to extract theprecious metals, is estimated to have released approximately156,000 Mg of mercury into the atmosphere and over 250,000 Mgof mercury into the environment between 1580 and 1900 (Nriagu,1994). The North American gold and silver rushes that began whengold was discovered in California in 1847 represent a particularlyintense period of mining. Silver production rose from 1.2 Mg yr�1 in1850 to 940 Mg yr�1 in 1880 (Bureau of the Census, 1989). From1850 to 1900 atmospheric mercury emissions from gold and silvermining in the United States averaged 780 Mg yr�1 (Nriagu, 1994),compared to present day U.S. anthropogenic emissions of approx-imately 100 Mg yr�1 (Pacyna et al., 2006). Because mercury is bothtoxic and persistent in the environment, this mercury release is stilla concern today. Sites contaminated with mercury during thehistoric gold rush continue to cause mercury contamination inCalifornia (Alpers et al., 2005) and Nevada (Wayne et al., 1996)watersheds.

: þ1 206 543 0308.Jaegle).

All rights reserved.

Sediment cores provide a record of changes in mercury depo-sition through history. Lake sediments suggest that modernmercury deposition in the northern hemisphere is 2–4 times largerthan the pre-industrial background value (Lindberg et al., 2007;Biester et al., 2007). For the historic North American gold rushperiod, however, the record is less clear. Schuster et al. (2002) founda factor of 5 enhancement in total mercury concentration anddeposition in a Wyoming glacier ice core layer corresponding to thelate 19th century, which they attribute to gold-mining emissions. Incontrast, Lamborg et al. (2002) did not find a clear mining signal inlake sediments from Nova Scotia and New Zealand, and suggestthat mercury from historic mining remained close to its sourcerather than being deposited globally.

Modeling studies provide insight into the relationship betweenmining emissions of mercury and the deposition changes recordedin sediment core record. Hudson et al. (1995) included mercuryemissions of 2200 Mg yr�1 at the peak of the North American goldrush in a pre-technological to modern box-model simulation of themercury cycle. Compared to lake sediments from the upper Mid-west United States (Swain et al., 1992), these mining emissionsresult in too large a peak in mercury deposition around 1880(Hudson et al., 1995). Pirrone et al. (1998) estimate that NorthAmerican mercury emissions peaked at 1708 Mg yr�1 in 1879

S. Strode et al. / Atmospheric Environment 43 (2009) 2012–2017 2013

mostly due to mining emissions, but do not find a correspondingenhancement in this period in sediment cores from the Great Lakes.

In this study, we estimate mercury emissions from the NorthAmerican gold rush era based on records of gold and silver produc-tion. We then use the GEOS-Chem global atmosphere–ocean–landmercury model to calculate the global impact of these emissions ondeposition. Finally, we examine the consistency between our esti-mates and historic core records from around the world.

2. Methods

2.1. Historic core records

Numerous studies have used lake sediment, peat bog, or icecores to interpret the history of mercury deposition and infer theanthropogenic enhancement ratio (ER) given by the ratio ofmodern to pre-industrial accumulation rates (ERmodern/p-i). Severaluncertainties are important in relating these records to atmo-spheric deposition, including variability in sedimentation rates(Engstrom and Wright, 1984) and sediment focusing (Perry et al.,2005), post-deposition mobility of mercury within the core (Gobeiland Cossa, 1993), and the contribution of mercury from thecatchment area rather than from atmospheric deposition (Swainet al., 1992). For this study, we select cores from the literature thatreport mercury as accumulation rate rather than concentration toreduce the influence of variable sediment flux, and we excludestudies that report significant post-depositional redistribution ofmercury. Since we are comparing with a global model, we use coresfrom areas where atmospheric deposition is expected to dominateover runoff from local sources.

The long lifetime of total gaseous mercury (>6 months) impliesthat large emissions from the North American gold rush mining erashould have affected deposition globally, and not just in NorthAmerica. We thus assemble a global dataset including sedimentcore records from North America (Engstrom and Swain, 1997;Fitzgerald et al., 2005; Kamman and Engstrom, 2002; Lamborget al., 2002; Landers et al., 1998, 2008; Lorey and Driscoll, 1999;Swain et al., 1992), South America (Lacerda et al., 1999), Greenland(Asmund and Nielsen, 2000; Bindler et al., 2001), Europe (Vertaet al., 1989), Siberia (Landers et al., 1998), and New Zealand (Lam-borg et al., 2002). For greater geographic coverage, we also includeobservations from an ice core in North America (Schuster et al.,2002), and two peat bog records in South America (Biester et al.,2002) and Europe (Roos-Barraclough and Shotyk, 2003).

As some of the cores do not extend back to pre-industrialtimes and some lack the temporal resolution to provide an 1880value, we define 3 enhancement ratios: ERmodern/p-i, ER1880/p-i,and ER1880/modern. ER1880/p-i is the product of the other two ratios.Modern is defined as the sediment data point closest to the year2000; p-i represents the pre-industrial period, which includesdata from before 1840; and 1880 is the date chosen to representthe North American gold and silver rush. If multiple cores lie inthe same model grid box or within 2 degrees latitude andlongitude of each other, we average them together. Table S1 inthe supplemental materials shows the dataset used for this study.

2.2. A priori mining emissions

We derive our global a priori mining emission inventory for the1870s using the following equation:

Fmining ¼�

Pgold þ Psilver

�� RHg=metal � fatmos (1)

where Fmining is the total mass of elemental mercury (Hg0) releasedto the atmosphere. Pgold is the mass of gold produced, Psilver is the

mass of silver produced, RHg/metal is the mass ratio of mercury lostto gold or silver produced and fatmos is the fraction of mercuryreleased to the atmosphere. Mitchell (2003a,b) reports the mass ofgold and silver produced by country for each year. Silver domi-nates the total production, with the United States producing760 Mg silver compared to 50 Mg gold in 1875 (Mitchell, 2003b).To determine Pgoldþ Psilver, we sum Mitchell’s (2003a,b) gold andsilver production numbers and average from 1870 to 1879 toaccount for the long atmospheric lifetime of Hg0 and the temporalresolution of the sediment cores. Mining emissions are distributedevenly across each country, except in the United States, where weassume that the emissions occurred in the western part of thecountry.

The ratio of mercury lost to gold or silver produced is uncertain.In 18th century South American silver mining, RHg/metal was esti-mated to be approximately 2:1, although in some regions the ratiowas closer to 1.5:1 or 1:1 and the ratio varied greatly betweenregions and years depending on the ore and the availability ofmercury (Brading and Cross, 1972; Fisher, 1977; Nriagu, 1994).Nriagu (1994) estimate that the ratio was approximately 1:1 in 19thcentury South and Central America. Pfeiffer and Lacerda (1988)estimated a ratio of 1.3:1 for modern gold miners using mercuryamalgamation in Brazil. Other modern estimates usually liebetween 1:1 and 1.5:1 (Lacerda, 2003). Given that the publishedestimates for the ratio of mercury lost to gold and silver overlap, wechoose a common value of RHg/metal¼ 1.5:1 for our a priori 1870semissions estimate.

Estimates also vary for the fraction of mercury released to theatmosphere during the amalgamation process, fatmos. Pfeiffer andLacerda (1988) estimated that 55% of mercury from modernAmazon gold mining enters the atmosphere, and Lacerda andSalomons (1991) report that 65%–83% is lost to the atmosphere inthis region. In contrast, Swain et al. (2007) estimate that only 30% ofthe mercury used in small-scale gold mining is released directly tothe atmosphere. For historic mining, Nriagu (1994) estimates a 60%loss to the atmosphere, and Pirrone et al. (1998) used this 60% valueto estimate emissions from North American gold and silver mining.Following these studies, we set fatmos to 0.6 for our a prioriemissions.

2.3. Global model

The GEOS-Chem chemical transport model (Bey et al., 2001)simulates mercury in the atmosphere–ocean–land system (Selinet al., 2007, 2008; Strode et al., 2007). The simulation includestracers for elemental mercury (Hg0), divalent mercury (HgII), andparticulate mercury (Hgp), with both oxidation of Hg0 to HgII andin-cloud reduction of HgII occurring in the atmosphere. We use heremodel version 7.04 (http://www.harvard.as.edu/chemistry/trop/geos) with updates described in Selin and Jacob (2008).

We conduct model simulations for 3 different sets of mercuryemissions: pre-industrial, 1880, and modern day. For each emis-sion scenario, we run the model until it reaches steady state. Themodel has a horizontal resolution of 4� latitude by 5� longitude,and 30 vertical levels. It is driven by assimilated meteorologicalfields from the NASA Goddard Earth Observing System (GEOS-4)for 2004 for all simulations so that the ER is not affected by inter-annual variability in precipitation. We compare the ER valuesfrom the cores with modeled deposition enhancement ratios forthe same time periods. Comparing ER values rather than absolutedeposition rates normalizes out some site-specific factors such asaverage local precipitation (Biester et al., 2007). Note that modernoxidant concentrations (OH and O3) were not modified for thepre-industrial and 1880 simulations. Thus, we do not address theeffects of changing oxidant concentrations on deposition.

Table 2Atmospheric mercury emissions from gold and silver mining for the United States,North America and the globe.

Years Emissions, Mg yr�1 Reference

United States North America Global

1800–1920 (25–1664) 396 (53–2331) Pirrone et al. (1998)1870–1880 1200 (700–1708) Pirrone et al. (1998)1850–1900 780 (208–1660) Nriagu (1994)1880 2200 Hudson et al. (1995)1870–1880 960 1350 1630 This study, a priori1870-1880 490

(200–810)660(270–1100)

820(330–1360)

This study, revised

Reported emissions for the various studies are given as mean values for the timeperiod, with the range of emissions indicated in parentheses.

S. Strode et al. / Atmospheric Environment 43 (2009) 2012–20172014

Selin et al. (2008) describes the pre-industrial and modernsimulations. Briefly, the pre-industrial simulation includes emis-sions of 1260 Mg yr�1 from the ocean and 1540 Mg yr�1 from land(Table 1). The ocean model simulates the coupled interactions ofthe mixed layer with the atmosphere and deep ocean, as well asconversion among three aqueous mercury species: elemental,divalent and non-reactive. The mechanistically parameterized landsource includes geogenic, evapotranspiration, and soil volatiliza-tion sources, and prompt recycling of deposition. The modernsimulation includes an additional 3400 Mg yr�1 anthropogenicsource and 660 Mg yr�1 biomass burning emissions. This increasein emissions leads to increasing deposition to land and oceansurfaces, and thus increases the cycling of mercury through thesereservoirs. With our interactive coupling of the atmosphere withthe land and ocean, we simulate increasing ocean and land emis-sions of 2960 Mg yr�1 and 2180 Mg yr�1, respectively (Table 1). Themodern mercury simulation has been validated against observa-tions of atmospheric surface concentrations, wet deposition overland, and oceanic aqueous concentrations, yielding a generallyunbiased simulation (Selin et al., 2007, 2008; Selin and Jacob, 2008;Strode et al., 2007, 2008).

For the 1880 simulation, we add our mining emission estimatefor the 1870s (Section 2.2) to the pre-industrial simulation of Selinet al. (2008). We will compare the results of this simulation to corerecords for 1880, because of the averaging time between input tothe lake and deposition in the sediments.

3. Results and discussion

3.1. A priori mining emissions

Based on equation (1), we obtain global mercury emissions frommining of 1630 Mg yr�1 for the 1870s. We assume that theseemissions occur as Hg0. For the United States, we find mercuryemissions from gold and silver mining in the 1870s of 960 Mg yr�1.Our a priori mining emissions for the United States lie in the middleof the range estimated by Nriagu (1994) and Pirrone et al. (1998)(Table 2). Considering all of North America, we estimate miningemissions of 1350 Mg yr�1, 83% of global mining emissions.

Pirrone et al. (1998) estimated North American mercury emis-sions ranged from under 750 Mg yr�1 in the early 1870s to over1500 Mg yr�1 during the late 1870s. For South America we estimatemercury emissions of 220 Mg yr�1, smaller than the 1821–1900average of 525 Mg yr�1 estimated by Nriagu (1994). In our inven-tory, mining emissions outside of America, occur in Japan, Australia,and New Zealand, and total 60 Mg yr�1.

Table 1Mercury budget for North America and the globe in Mg yr�1.

North Americaa Global

Pre-industrial 1880 Modern Pre-industrial 1880 Modern

Emissions (Mg yr�1)Anthropogenicb 0 660 180 0 820 3400Biomass burning 0 0 30 0 0 660Land 330 340 390 1540 1580 2180Ocean – – – 1260 1400 2960

Total Emissions 330 1000 600 2800 3800 9200

Deposition(wetþ dry)

180 300 570 2800 3800 9200

a Emissions and deposition for North America are only over land.b Includes direct emissions from mining and other anthropogenic activities. The

1880 mining emissions are for our revised inventory (Section 3.3).

3.2. Modeled and observed ERs

Before constraining the 1870s mining emissions, we examinethe ability of the model to reproduce the observed ERmodern/p-i.Selin et al. (2008) found good agreement between the modeled ERand core records from the upper Midwest U.S. and New Zealand.Fig. 1 compares the modeled ERmodern/p-i with a more extensivedataset of core records, described in Section 2.1 and Table S1.

The model shows good agreement with core records from areassuch as the upper Midwest U.S. (observations: 3.2, model: 3.2) andChile (observations: 2.6, model: 2.9). It also captures the greateranthropogenic impact in industrialized regions such as the north-east U.S. and Europe. The model greatly underestimates theWyoming ice core value of 11, but this value is much higher thanthe values from lake sediments (1–4) and may relate to the diffi-culty in interpreting the ice core record. In addition, the model doesnot capture much of the site-to-site variability seen in the cores.This may be due to local sources not captured by the model oruncertainty in the individual core records. The model is also unableto reproduce the low enhancement ratios at high latitudes found byLanders et al. (1998) in Siberia (1.1–1.3) and Alaska (1.0–1.3). Themodel does reproduces the Alaskan observations reported byFitzgerald et al. (2005) (observations¼ 3.2, model¼ 2.8). A possibleexplanation for the low ER values at some high latitude sites isa large mercury source from erosion of naturally enriched soils. Thislarger background input would reduce the relative impact ofatmospheric deposition and thus reduce the observed modern topre-industrial enhancement ratio (Fitzgerald et al., 2005). Another

Fig. 1. Modeled ratio of modern to pre-industrial deposition (ERmodern/p-i). Corerecords are shown by the circles, which are color coded according to the observedERmodern/p-i.

Table 3Deposition enhancement ratios from historic cores and model for modern/pre-industrial, 1880/pre-industrial and 1880/modern.

Observations(mean� s)

A priori model Revised model

Sites Global Modelbias

Sites Global ModelBias

ERmodern/p-i 2.7� 1.4 3.1� 0.3 3.1� 0.7 þ51% – – –ER1880/p-i 1.4� 0.3 1.8� 0.2 1.7� 0.2 þ29% 1.4� 0.1 1.3� 0.1 0%ER1880/modern 0.5� 0.2 0.6� 0.2 0.5� 0.1 þ39% 0.5� 0.1 0.4� 0.1 þ4%

Model values (mean�s) are presented for the core site locations and the entireworld. The mean model bias is calculated for the core sites.

S. Strode et al. / Atmospheric Environment 43 (2009) 2012–2017 2015

possible explanation is differences in the time period considered‘‘pre-industrial’’ for the different studies.

Table 3 presents a comparison of observed and modeled ER’s.The Wyoming ice core is excluded from the calculations, as it is anoutlier compared to other observations (Table S1). The mean (�standard deviation) observed ERmodern/p-i from 18 locations is2.7�1.4 (Table 3), while the corresponding model ERmodern/p-i is3.1�0.3. The model overestimate of the Landers et al. (1998) sitesin Siberia and Alaska drives the positive model bias of 51% shown inTable 3 (model bias is defined as the mean normalized differencebetween model and observations: [model-observations]/observa-tions). Removing these sites increases the observed mean ratio andreduces the bias in ERmodern/p-i to �3%.

Fig. 2 (left) shows the global distribution of modeled ER1880/p-i.ER1880/p-i is largest over the western U.S. and Mexico, where itexceeds 2 due to large mining emissions during the 1870s inthese regions. There is an inter-hemispheric gradient due to thelarger increase in source strength in the northern hemisphere.The mean modeled ER1880/p-i is 1.8 in the northern hemisphereand 1.6 in the southern hemisphere. In North America, the meanobserved ER1880/p-i is 1.4� 0.3 while the mean model ER1880/p-i is2.0� 0.4. The model overestimates ER1880/p-i at remote sites inAlaska and New Zealand. This model overestimate suggests thatmining emissions may be overestimated. An exception is theWyoming ice core, which predicts an ER1880/p-i of 5 for the goldrush era compared to the modeled value of 2.7. The model alsounder predicts the ER from the lake core record from CarajasMountain, Brazil (Lacerda et al., 1999). This may be due to a localsource not included in our model emissions.

Fig. 2 (right) shows the modeled and observed ER1880/modern.Small ER1880/modern values in East Asia reflect the large modernanthropogenic source in this region. The model generally over-predicts ER1880/modern, particularly in the western U.S. The observedmean ER1880/modern in the western U.S. is 0.4 while the mean model

Fig. 2. Modeled ER1880/p-i (left) and ER1880/modern (right) with a priori mining em

ER1880/modern is 0.8. Table 3 shows a 29% positive mean model biasin ER1880/p-i and a 39% positive bias in ER1880/modern, both suggestingthat mining emissions should be reduced.

3.3. Revised 1870s mining emissions

Given the disagreement between the observed and modeledER’s for the gold rush period, we derive a revised mining emissionsestimate to bring the model into better agreement with the globalcore record. Reducing our global a priori mining emissions bya factor of 2 from 1630 Mg yr�1 to 820 Mg yr�1 removes the modelbias in ER1880/p-i (Table 3) (minimizing the bias for ER1880/modern

requires reduction by a factor of 2.3, yielding mining emissions of710 Mg yr�1). Given the uncertainties in both the model and thehistoric core records, we determine an uncertainty range for themining emissions by using the GEOS-Chem model to calculateemissions consistent with the observed mean ER1880/p-i� onestandard deviation (ER1880/p-i¼ 1.1–1.7). This yields estimates ofmining emissions ranging between 330 and 1360 Mg yr�1 (Table 2).

Fig. 3 maps ER1880/p-i and ER1880/modern using the revised miningestimate of 820 Mg yr�1, and displays improved agreement withobservations. In particular, the modeled ER1880/modern over NorthAmerica is reduced from 2.0� 0.4 to 1.5� 0.2, and compares wellwith observations (1.4� 0.3). Based on these revised global miningemissions, we estimate that gold and silver mining in the UnitedStates released 490 Mg yr�1 of mercury to the atmosphere with anuncertainty range of 200–810 Mg yr�1 (Table 2).

Assuming that the production of gold and silver is relativelywell constrained, equation (1) implies that a reduction in mercuryreleased to the atmosphere can result from a reduction in RHg/metal

and/or fatmos. A reduction in RHg/metal implies lower total mercuryrelease to the environment, while a reduction in fatmos implies thatmore mercury was deposited to local soil and watersheds whileless was exported to the global atmosphere. If our fatmos value of0.6 is correct, then to obtain a 50% reduction in mining emissions,RHg/metal must be reduced from 1.5 to 0.75, lower than mostpublished estimates. Alternatively, if the RHg/metal value is correct,fatmos must be reduced from 0.6 to 0.3, which is within publishedestimates, albeit at the low end. The total reduction could also beobtained by smaller reductions in both RHg/metal and fatmos.

3.4. Mercury budget for 1880

Atmospheric emissions from gold and silver mining hada significant impact on global deposition during the late 19thcentury. Table 1 compares the mercury budget for the 1880

issions of 1630 Mg yr�1. Observations are shown by the color-coded circles.

Fig. 3. Same as Fig. 2, but with the revised mining emissions (820 Mg yr�1).

S. Strode et al. / Atmospheric Environment 43 (2009) 2012–20172016

simulation (with revised mining emissions) to the modern and pre-industrial simulations. Both the global and North American budgetsare summarized. North American mercury emissions more thantripled between the pre-industrial period and 1880, increasingfrom 330 to 1000 Mg yr�1. Globally, emissions increased by 35%(Table 1). The resulting deposition increased by 60% (30%) overNorth America (globally) in 1880.

Between 1880 and present time, North American emissions(including biomass burning and land emissions) decreased by 40%from 1000 to 600 Mg yr�1. Over that same time period, however,deposition to North America nearly doubled. This is due to thefactor of 2.4 increase in global emissions in the modern worldcompared to 1880, and reflects the fact that a large fraction ofmodern deposition over North America is due to the oxidation andscavenging of Hg0 from the global pool (Selin and Jacob, 2008).

Globally, our 1880s emissions of 3800 Mg yr�1 include a directmining source of 820 Mg yr�1 and emissions from land and oceanof 1580 Mg yr�1 and 1400 Mg yr�1, respectively. Relative to the pre-industrial simulation, emissions from these two reservoirs areenhanced by a total of 170 Mg yr�1 because of the recycling ofdeposited mining emissions.

Fig. 4 summarizes calculated ERmodern/p-i and ER1880/p-i for eachcontinent. While the majority of the mining emissions occurred in

Fig. 4. Deposition enhancement ratio spatially averaged over each continent and thewhole world. Light gray bars represent ERmodern/p-i and dark gray bars representER1880/p-i. The 1880 simulation uses the revised mining estimate. The symbols showthe mean observed ratios from North American and global cores for ERmodern/p-i

(circles) and ERmodern/p-i (triangles) with error bars representing the standarddeviation. The filled black circles show the mean observed ERmodern/p-i for all coresexcept the Wyoming ice core, while the open circles show ERmodern/p-i also excludeLanders et al. (1998) observations in Alaska and Siberia as discussed in the text.

North America, the impact on deposition is spread across allcontinents because of the long lifetime of atmospheric Hg0.ERmodern/p-i shows greater variability between continents thanER1880/p-i in part because modern emissions occur not only as Hg0,but also as HgII and Hgp, which can be deposited locally andregionally. For North America and the global average, the modeledratios are close to the observed mean. Observations on othercontinents are too sparse to calculate a continental average.

Several key uncertainties are important in the mining emissionestimates. We have assumed no anthropogenic emissions for the1880 simulation, while in reality the industrial revolution hadstarted by this point. Consequently, our mining emission estimatesshould be viewed as an upper limit on the true mining source. Pre-industrial land and ocean emissions are also uncertain since thereare no direct measurements of these fluxes. While the modelreproduces well the mean ER values of the core record, cores takenfrom nearby locations can show substantial variability not capturedby the model. The variability between cores may be due to eitherlocal processes or uncertainties in the interpretation of the cores. Tobetter constrain historic mercury fluxes, a greater number of coresfrom remote regions outside of North America would be veryuseful.

4. Conclusions

We estimate atmospheric mercury emissions from gold andsilver mining during the peak of the North American gold rush. Oura priori mining emissions estimate is based on gold and silverproduction averaged over the 1870s and multiplied by the ratio ofmercury lost to gold or silver produced and the fraction of mercuryemitted to the atmosphere. We include these emissions in a simu-lation of the global mercury cycle for the year 1880.

The modeled enhancement in mercury deposition between thepre-industrial era and present compares well with the meanenhancement seen in sediment core records, although it missesmuch of the observed variability. However, the modeled enhance-ment due to mining emissions in the 1880 simulation over-estimates the observed enhancement with a mean bias of 29%. Toimprove the consistency with observations, we revise our estimateof 1870s mining emissions of atmospheric mercury downwardfrom 1630 Mg yr�1 to 820 Mg yr�1. Globally, this leads to a deposi-tion enhancement ratio of 1.3 for the 1880 versus pre-industrialsimulations.

Lower atmospheric emissions from 1870s mining implya smaller impact of the North American gold rush on globalmercury deposition. However, if a smaller fraction of the mercuryused in gold and silver mining was emitted to the global

S. Strode et al. / Atmospheric Environment 43 (2009) 2012–2017 2017

atmosphere, then a larger fraction may have been deposited locallyto water and soil. More sediment cores from remote regionsthroughout the world would be valuable for reducing the uncer-tainty in the global impact of historic gold and silver mining onmercury deposition.

Acknowledgements

This work was supported by funding from the National ScienceFoundation under grant ATM 0238530.

Appendix. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.atmosenv.2009.01.006.

References

Alpers, C.N., Hunerlach, M.P., May, J.T., Hothem, R.L., Taylor, H.E., Antweiler, R.C., DeWild, J.F., Lawler, D.A., 2005. Geochemical characterization of water, sediment,and biota affected by mercury contamination and acidic drainage from histor-ical gold mining, Greenhorn Creek, Nevada County, California, 1999–2001. U.S.Geological survey scientific investigations report 2004-5251. Available from:http://pubs.usgs.gov/sir/2004/5251.

Asmund, G., Nielsen, S.P., 2000. Mercury in dated Greenland marine sediments.Science of the Total Environment 245, 61–72.

Bey, I., Jacob, D.J., Yantosca, R.M., Logan, J.A., Field, B.D., Fiore, A.M., Li, Q.B.,Liu, H.G.Y., Mickley, L.J., Schultz, M.G., 2001. Global modeling of troposphericchemistry with assimilated meteorology: model description and evaluation.Journal of Geophysical ResearchdAtmospheres 106 (D19), 23073–23095.

Biester, H., Bindler, R., Martinez-Cortizas, A., Engstrom, D.R., 2007. Modeling thepast atmospheric deposition of mercury using natural archives. EnvironmentalScience and Technology 41 (14), 4851–4860.

Biester, H., Kilian, R., Franzen, C., Woda, C., Mangini, A., Sholer, H.F., 2002. Elevatedmercury accumulation in a peat bog of the Magellanic Moorlands, Chile (53�S) –an anthropogenic signal from the Southern Hemisphere. Earth and PlanetaryScience Letters 201, 609–620.

Bindler, R., Renberg, I., Appleby, P.G., Anderson, N.J., Rose, N.L., 2001. Mercuryaccumulation rates and spatial patterns in lake sediments from WestGreenland: a coast to ice margin transect. Environmental Science and Tech-nology 35, 1736–1741.

Brading, D.A., Cross, H.E., 1972. Colonial silver mining: Mexico and Peru. HispanicAmerica Historical Review 52 (4), 545–579.

Bureau of the Census, 1989. Historical Statistics of the Unites States: Colonial Timesto 1970. U.S. Department of Commerce, Kraus International Publications, WhitePlains, New York.

Engstrom, D.R., Wright Jr., H.E., 1984. Chemical stratigraphy of lake sediments asa record of environmental change. In: Haworth, E.Y., Lund, J.G. (Eds.), LakeSediments and Environmental History. University of Minnesota Press, Minne-apolis, MN.

Engstrom, D.R., Swain, E.B., 1997. Recent declines in atmospheric mercury depositionin the upper midwest. Environmental Science and Technology 31 (4), 960–967.

Fisher, J.R., 1977. Silver Mines and Silver Miners in Colonial Peru, 1776–1824. Centerfor Latin American Studies, University of Liverpool, Liverpool.

Fitzgerald, W.F., Engstrom, D.R., Lamborg, C.H., Tseng, C.-M., Balcom, P.H.,Hammerschmidt, C.R., 2005. Modern and historic atmospheric mercury fluxesin northern Alaska: global sources and arctic depletion. Environmental Scienceand Technology 39, 557–568.

Gobeil, C., Cossa, D., 1993. Mercury in sediments and sediment pore water in the Lau-rentian Trough. Canadian Journal of Fisheries and Aquatic Sciences 50, 1794–1800.

Hudson, R.J.M., Gherini, S.A., Fitzgerald, W.F., Porcella, D.B., 1995. Anthropogenicinfluences on the global mercury cycle: a model-based analysis. Water, Air, andSoil Pollution 80, 265–272.

Kamman, N.C., Engstrom, D.R., 2002. Historical and present fluxes of mercury toVermont and New Hampshire lakes inferred from 210Pb dated sediment cores.Atmospheric Environment 36, 1599–1609.

Lacerda, L.D., Salomons, W., 1991. Mercury in the Amazon: A Chemical Time Bomb?Dutch Ministry of Housing. Physical Planning and Environment Report Haren.

Lacerda, L.D., Ribeiro, M.G., Cordeiro, R.C., Sifeddine, A., Turcq, B., 1999. Atmosphericmercury deposition over Brazil during the past 30,000 years. EnvironmentalBiodiversity 51 (5/6), 363–371.

Lacerda, L.D., 2003. Updating global Hg emissions from small-scale gold mining andassessing its environmental impacts. Environmental Geology 43, 308–314.

Lamborg, C.H., Fitzgerald, W.F., Damman, A.W.H., Benoit, J.M., Balcom, P.H.,Engstrom, D.R., 2002. Modern and historic atmospheric mercury fluxes in bothhemispheres: global and regional mercury cycling implications. GlobalBiogeochemical Cycles 16 (4), 1104. doi:10.1029/2001GB001847.

Landers, D.H., Gubala, C., Verta, M., Lucotte, M., Johansson, K., Vlasova, T.,Lockhart, W.L., 1998. Using lake sediment mercury flux ratios to evaluate theregional and continental dimensions of mercury deposition in arctic and borealecosystems. Atmospheric Environment 32 (5), 919–928.

Landers, D.H., Simonich, S., Jaffe, D., Geiser, L., Campbell, D.H., Schwindt, A.,Schreck, C., Kent, M., Hafner, W., Taylor, H.E., Hageman, K., Usenko, S.,Ackerman, L., Schrlau, J., Rose, N., Blett, T., Erway, M.M., 2008. The fate, trans-port, and ecological impacts of airborne contaminants in western nationalparks (USA). In: Christie, S. (Ed.), Western Airborne Contaminants AssessmentProject Final Report, vol. 1 EPA/600/R-07/138.

Lindberg, S., Bullock, R., Ebinghaus, R., Engstrom, D., Feng, X., Fitzgerald, W.,Pirrone, N., Prestbo, E., Seigneur, C., 2007. A synthesis of progress and uncer-tainties in attributing the sources of mercury in deposition. Ambio 36, 19–32.

Lorey, P., Driscoll, C.T., 1999. Historical trends of mercury deposition in Adirondacklakes. Environmental Science and Technology 33, 718–722.

Mitchell, B.R., 2003a. International Historical Statistics: Africa, Asia, and Oceania1750–2000. Palgrave MacMillan, New York.

Mitchell, B.R., 2003b. International Historical Statistics: The Americas 1750–2000.Palgrave MacMillan, New York.

Nriagu, J.O., 1994. Mercury pollution from the past mining of gold and silver in theAmericas. Science of the Total Environment 149, 167–181.

Pacyna, E.G., Pacyna, J.M., Steenhuisen, F., Wilson, S., 2006. Global anthropogenicmercury emission inventory for 2000. Atmospheric Environment 40 (22),4048–4063.

Perry, E., Norton, S.A., Kamman, N.C., Lorey, P.M., Driscoll, C.T., 2005. Deconstructionof historic mercury accumulation in lake sediments, northeastern United States.Ecotoxicology 14, 85–99.

Pfeiffer, W.C., Lacerda, L.D., 1988. Mercury inputs into the Amazon region, Brazil.Environmental Technology Letters 9, 325–330.

Pirrone, N., Allegrini, I., Keeler, G.J., Nriagu, J.O., Rossmann, R., Robbins, J.A., 1998.Historic atmospheric mercury emissions and depositions in North Americacompared to mercury accumulations in sedimentary records. AtmosphericEnvironment 32 (5), 929–940.

Roos-Barraclough, F., Shotyk, W., 2003. Millennial-scale records of atmo-spheric deposition obtained from ombotrophic and minerotrophic peat-lands in the Swiss Jura Mountains. Environmental Science and Technology37, 235–244.

Schuster, P.F., Krabbenhoft, D.P., Naftz, D.L., Cecil, L.D., Olson, M.L., Dewild, J.F.,Susong, D.D., Green, J.R., Abbott, M.L., 2002. Atmospheric mercury depositionduring the last 270 years: a glacial ice core record of natural and anthropogenicsources. Environmental Science and Technology 36, 2303–2310.

Selin, N.E., Jacob, D.J., Park, R.J., Yantosca, R.M., Strode, S., Jaegle, L., Jaffe, D.A., 2007.Chemical cycling and deposition of atmospheric mercury: global constraintsfrom observations. Journal of Geophysical Research 112, D02308. doi:10.1029/2006JD007450.

Selin, N.E., Jacob, D.J., Yantosca, R.M., Strode, S., Jaegle, L., Sunderland, E.M., 2008.Global 3-D land–ocean–atmosphere model for mercury: present-day vs pre-industrial cycles and anthropogenic enrichment factors for deposition. GlobalBiogeochemical Cycles 22, GB2011. doi:10.1029/2007GB003040.

Selin, N.E., Jacob, D.J., 2008. Seasonal and spatial patterns of mercury wet depositionin the United States: constraints on the contribution from North Americananthropogenic sources. Atmospheric Environment 42, 5193–5204. doi:10.1016/j.atmosenv.2008.02.069.

Strode, S., Jaegle, L., Selin, N.E., Jacob, D.J., Park, R.J., Yantosca, R.M., Mason, R.P.,Slemr, F., 2007. Global simulation of air–sea exchange of mercury. GlobalBiogeochemical Cycles 21, GB1017.

Strode, S., Jaegle, L., Jaffe, D.A., Swartzendruber, P.C., Selin, N.E., Holmes, C., Yantosca,R.M., 2008. Trans-pacific transport of mercury. Journal of Geophysical Research-Atmospheres 113, D15305.

Swain, E.B., Engstrom, D.R., Brigham, M.E., Henning, T.A., Brezorik, P.L., 1992.Increasing rates of atmospheric mercury deposition in midcontinental NorthAmerica. Science 257, 784–787.

Swain, E.B., Jakus, P.M., Rice, G., Lupi, F., Maxson, P.A., Pacyna, J.M., Penn, A.,Spiegel, S.J., Veiga, M.M., 2007. Socioeconomic consequences of mercury useand pollution. Ambio 36 (1), 45–61.

Verta, M., Tolonen, K., Simola, H., 1989. History of heavy metal pollution in Finlandas recorded by lake sediments. Science of the Total Environment 87/88, 1–18.

Wayne, D.M., Warwick, J.J., Lechler, P.J., Gill, G.A., Lyons, W.B., 1996. Mercurycontamination in the Carson river, Nevada: a preliminary study of the impact ofmining wastes. Water, Air, and Soil Pollution 92, 391–408.