For Peer Review Lead Pollution Resulting from Roman Gold Extraction in Northwestern Spain Journal: The Holocene Manuscript ID HOL-16-0185.R1 Manuscript Type: Paper Date Submitted by the Author: n/a Complete List of Authors: Hillman, Aubrey; University of Louisiana at Lafayette, Environmental Science Abbott, Mark; University of Pittsburgh, Geology and Planetary Science Valero-Garc!es, B.L.; Instituto Pirenaico de Ecologia - CSIC, Geoenvironmental Processes and Global Change Morellon, Mario; Universidad de Cantabria, CITIMAC, Facultad de Ciencias Barreiro-Lostres, Fernando; Instituto Pirenaico de Ecología - CSIC, Procesos Geoambientales y Cambio Global Bain, Daniel; University of Pittsburgh, Keywords: Spain, late Holocene, lead pollution, Las Médulas, Romans, paleolimnology Abstract: Roman mining and metallurgy left a detectable signal of lead pollution throughout Europe, northern Africa, and the Middle East. Las Médulas, in Northwestern Iberia, was the largest Roman gold mine and fundamentally altered the local landscape. To document the environmental consequences of this activity, we present a 4000 year record of lake sediment geochemistry from Laguna Roya, 35 km south of Las Médulas. Using the concentrations of trace metals weakly bound to sediment including lead, antimony, bismuth, and arsenic, we find increased levels of these metals from 400 300 BC to 210 120 AD, during the Roman Republic/Empire. We attribute these increases to the atmospheric deposition of heavy metals arising from the regional extraction, processing, and/or smelting of gold ores. Lead pollution at the peak of this activity (170 15 BC) is twice as high as modern-day concentrations, suggesting that the amount of pollution generated by pre-Industrial civilizations is much larger than previously estimated. We find additional increases in antimony and bismuth from 1500 to 1800 1700 AD, possibly associated with post- medieval mining activity. Concentrations of lead begin to increase again ~1850 1860 AD during the start of the Industrial Revolution and reach a peak at 1980 1990 AD. Declining modern-day levels of lead can be attributed to the phase out of leaded gasoline. This is one of only a handful of studies to document pre-industrial pollution levels substantially higher than present-day, adding to a growing body of evidence that anthropogenic environmental degradation has been taking place for several thousands of years. http://mc.manuscriptcentral.com/holocene HOLOCENE For Peer Review Page 1 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 1 ABSTRACT 1 Roman mining and metallurgy left a detectable signal of lead pollution throughout 2 Europe, northern Africa, and the Middle East. Las Médulas, in Northwestern Iberia, was the 3 largest Roman gold mine and fundamentally altered the local landscape. To document the 4 environmental consequences of this activity, we present a 4000 4000-year record of lake 5 sediment geochemistry from Laguna Roya, 35 km south of Las Médulas. Using the 6 concentrations of trace metals weakly bound to sediment including lead, antimony, bismuth, and 7 arsenic, we find increased levels of these metals from 400 300 BC to 210 120 AD, during the 8 Roman Republic/Empire. We attribute these increases to the atmospheric deposition of heavy 9 metals arising from the regional extraction, processing, and/or smelting of gold ores. Lead 10 pollution at the peak of this activity (170 15 BC) is twice as high as modern-day concentrations, 11 suggesting that the amount of pollution generated by pre-Industrial civilizations and the 12 associated environmental impacts areis much larger than previously estimated. We find 13 additional increases in antimony and bismuth from 1500 to 1800 1700 AD, possibly associated 14 with post-medieval mining activity. Concentrations of lead begin to increase again ~1850 1860 15 AD during the start of the Industrial Revolution and reach a peak at in 1980 1990 AD. Declining 16 modern-day levels of lead can be attributed to the phase out of leaded gasoline. This is one of 17 only a handful of studies to document pre-industrial pollution levels substantially higher than 18 present-day, adding to a growing body of evidence that anthropogenic environmental 19 degradation has been taking place for several thousands of years. 20 21 22 23 Page 2 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 2 Introduction 24 Northwestern Iberia is rich in mineral resources, particularly economically extractable 25 and relatively abundant gold deposits. These resources have been exploited for several 26 millennia, most extensively during the Roman period from the first to the third centuries AD 27 (Center, 2015). During the height of mineral resource exploitation, an estimated 600 million 28 cubic meters of earth were moved (Pérez-García et al., 2000) and around 195 tons of gold were 29 extracted (Gómez-Fernández et al., 2012). Most of this extraction was from alluvial gold 30 deposits, however mesothermal gold bearing quartz veins are abundant in this region as well 31 (Gómez-Fernández et al., 2012) (Fig. 1) and were also heavily exploited. Nearly 15% of all 32 Roman gold exploitation took place at the mining area known as Las Médulas, a World Heritage 33 Landscape Site since 1997. Las Médulas contains gold bearing gravel, overlain by extensive 34 alluvial fan deposits (Pérez-García et al., 2000). Extraction relied on a technique known as ruina 35 montium where water was forced through large pits in the alluvial deposits to cause wholescale 36 collapse and subsequent hydraulic sorting of the gold bearing gravel (Ruiz del Árbol Moro et al., 37 2014). Additional extraction of ores from hard-rock deposits relied on the technique of fire-38 setting where fires were set against stone and rapidly quenched, causing the rock face to shatter 39 (Weisgerber and Willies, 2000). This method was extensively practiced on the gold deposits of 40 northwestern Spain (Pliny, 1952). All of these processes fundamentally transformed the 41 landscape and these transformations persist to present. 42 Ores that were mined were eventually purified by the cupellation method where 43 impurities were driven off in a high temperature furnace (Healy, 1978). These impurities, such 44 as lead, were volatized, transported atmospherically, and eventually deposited as wet and dry 45 fallout over the landscape. Lead from Iberia has been detected in Greenland ice cores using lead 46 Page 3 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 3 isotopes as tracers (Hong et al., 1994) and is so widespread and concurrent in European lakes and 47 peat bogs as be suggested as a chronological marker for the Roman Period (Renberg et al., 48 2001). However, some differences in the magnitude and timing of increases in lead are 49 observed, demonstrating the unique regional histories of mineral resource extraction. For 50 example, both Zoñar (Fig. 1) (Martín-Puertas et al., 2010) and Río de Seco (García-Alix et al., 51 2013) lakes near the Strait of Gibraltarin Southern Spain record increases in lead associated with 52 Roman metallurgy as well as earlier and larger increases. Both the Phoenicians and 53 Carthaginians exploited Río Tinto in southwestern Spain, an area rich in silver (Richardson, 54 1976). These mining activities left behind sediments contaminated in heavy metals (Leblanc et 55 al., 2000) and the lead signature in the Greenland ice cores has been attributed, in part, to 56 Carthaginian mines as well as Roman ones (Rosman et al., 1997). Most importantly for this 57 study, the initiation of large-scale mineral resource exploitation at Las Médulas, by Roman and 58 earlier cultures, remains uncertain (Lewis and Jones, 1970). 59 Most studies documenting Roman lead pollution have been focused on northern Europe 60 (Brannvall et al., 1997; Mighall et al., 2009; Shotyk et al., 1998), though a few studies have 61 looked at this disturbance in northern Spain (Fig. 1). Molina mire in NW Spain found evidence 62 of Roman metallurgy as well as two hundred more years of additional mining activity after the 63 decline of the Roman Empire (Martínez-Cortizas et al., 2013). The peak of metallurgical activity 64 was accompanied by shifts in vegetation communities as well as changes in the hydrologic 65 balance of the bog, likely due to human manipulation of water flow associated with ore washing 66 (López-Merino et al., 2014). Another study of a more westerly peat bog, Penido Vello, also 67 found increases in lead associated with Roman metallurgical activity (Martínez-Cortizas et al., 68 1997), and a follow-up study used lead isotopes and was able to source the Roman pollution to 69 Page 4 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 4 NW Spanish ores (Kylander et al., 2005), but did not find evidence for post-Roman extended 70 metallurgical activity. An additional study of Penido Vello as well as several other peat bogs 71 from this region (Borralleiras da Cal Grande and Pena da Cadela) examined nickel, zinc, arsenic, 72 and cadmium and found concentration profiles similar to that of lead, peaking from 100 to 200 73 AD (Pontevedra-Pombal et al., 2013). In all of the aforementioned studies, pollution arising 74 from Roman metallurgical activities was clearly detectable, but below that of modern-day levels. 75 Redo Lake (Fig. 1) in the Pyrenees is unique in this regard. Pre-industrial lead 76 concentrations in these sediments were at levels twice that of the last 50 years and 14 times 77 higher than background concentration values at the peak in 650 AD (Camarero et al., 1998). Not 78 only is the magnitude of these lead concentrations surprising, but the timing is later than would 79 be expected. This suggests that either: 1) Roman mining and metalworking activity did not leave 80 a substantial signature in the lake and instead, this pollution was the result of medieval activities; 81 or 2) the timing of this pollution was not properly dated. Therefore, the timing of the increase in 82 lead remains uncertain and so several questions remainthe primary aim of this study is to clarify 83 regarding the magnitude and extent of post-Roman ore exploitation and mineral resource use not 84 just in northern Spain, but throughout Europe (Tylecote, 1992). 85 From an environmental perspective, the long history of metallurgical activities 86 contributed trace metal loadings to the landscape through wet and dry deposition of airborne 87 pollution. A study of soils from the central Pyrenees found that the remobilization of sediments 88 and soils contaminated by historical activities was a significant contributor to modern-day 89 pollution accumulation (Bacardit et al., 2012). Therefore, a second aim of this study is to clarify 90 metal distribution from this historical contamination that may be stored on the landscape of 91 Page 5 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 5 northwestern Spain, given that this is area has experienced substantial metallurgical activity over 92 several millennia. 93 To examine these questions, we present a 4000 year lake sediment record from Laguna 94 Roya, a lake that has ideal characteristics for recording atmospheric pollution. Metals weakly 95 sorbed to lake sediments have been used throughout Europe (Bindler et al., 2012; Brannvall et 96 al., 2001), South America (Abbott and Wolfe, 2003; Cooke et al., 2007), North America 97 (Pompeani et al., 2013), and Asia (Hillman et al., 2015; Lee et al., 2008) to infer metallurgical 98 activities proximal to the lake as well as long-range atmospheric pollution and provide the 99 conceptual basis for our own study. While we focus our attention on lead because it is relatively 100 immobile once sorbed to lake sediments (Gallon et al., 2004) and insensitive to water chemistry 101 changes such as oxidation or reduction potential (Hamilton-Taylor and Davison, 1978), we 102 supplement our interpretation with other metals including antimony, bismuth, and arsenic. 103 104 Setting 105 Laguna Roya (42°13’N, 6°46’W, 1608 m above sea level) is a small glacial lake in 106 northwest Spain with a maximum depth of 6.5 m and a surface area of 0.025 km2. A survey of 107 water quality observed a conductivity of 19 µS/cm, a pH of 7.3, and an alkalinity of 0.13 mEq/L 108 (Aldasoro et al., 1996). The catchment area is small (0.150 km2) and is primarily composed of 109 augen gneiss with a few diorite intrusions (Castro et al., 2003), with thinly developed soils (The 110 European Soil Database, 2004Askoy et al., 2010), and low angled slopes. Vegetation is sparse 111 and consists mainly of shrubs. There is limited hydrologic inflow and outflow, making it 112 particularly sensitive to atmospheric processes. An average of 1.6 m of precipitation falls 113 annually, primarily during the months of October to January (IAEA/WMO, 2014) and the 114 Page 6 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 6 predominant wind direction is from the southwest. While many lakes in this region of Spain 115 have been dammed, Roya has never been hydrologically modified due to its small size (Allen et 116 al., 1996) and today, there is minimal human occupation of the watershed. Las Médulas, the 117 largest Roman gold extraction operation, is 35 km north of Laguna Roya along with hundreds of 118 other alluvial gold deposits as well as gold veins within solid rock (Fig. 1). 119 Previous work on Laguna Roya focused on palynology (Allen et al., 1996). 120 Approximately 8.2 m of sediment were collected and an age model was developed on the basis 121 of Accelerator Mass Spectrometer (AMS) radiocarbon dates primarily on moss. Over 15,300 122 years, there were substantial changes in pollen reflecting temperature and moisture variations, 123 but most importantly for our own study, indicators of human disturbance within the watershed 124 were noted at a depth of around 184 cm, corresponding to 2700 years BP or 750 BC. 125 Anthropogenically driven vegetation changes around the watershed were inferred from the 126 presence of disturbance taxa such as Juglans (walnut), Castanea (chestnut), Cannabaceae 127 (hemp), and a variety of cultivation cereals. More recent work on Roya lake sediments focused 128 on higher resolution palynological and diatom assemblages as well as chironomid inferred July 129 air temperatures only from the period of 15,500 to 11,200 years BP (Muñoz Sobrino et al., 130 2013). In general, this work supported the findings of Allen et al., (1996), but did not focus on 131 the time period relevant for our own study. 132 133 Methods 134 Field Work 135 In 2014, a 1.24 m surface core with an intact sediment-water interface was collected 136 using a light weight percussion coring system from the deepest part of the lake at a water depth 137 Page 7 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 7 of 6.6 m. The upper 15 cm were sliced in the field at 0.5 cm intervals and used for geochemical 138 analysis and 210Pb dating. The sediment was sampled using plastic instruments and divided into 139 plastic bags, with little to no handling. Deeper sediment cores were collected from the same 140 location using a steel barrel Livingston corer (Wright et al., 1984) for a total of ten overlapping 141 drives. Overlapping cores were correlated on the basis of field notes, visible sedimentology, and 142 geochemical profiles to form a composite record of 6.96 m. 143 144 Water Content, Bulk Density, and Loss-On-Ignition Analysis 145 Water content, bulk density, and loss on ignition were measured at 2-cm intervals using 1 146 cm3 samples. Samples were dried at 60°C for 48 hours to remove water. Weight percent organic 147 matter and carbonate content was determined by loss-on-ignition at 550°C and 1000°C, 148 respectively (Dean, 1974). 149 150 Geochronology 151 Thirteen Sixteen AMS radiocarbon ages were measured on wood, charcoal, and plant 152 macrofossils (Table 1). These samples were pretreated using a standard acid, alkali, acid 153 procedure (Abbott and Stafford, 1996), measured at the Keck Center for Accelerator Mass 154 Spectrometry at the University of California Irvine, and calibrated using Calib 7.0 (Reimer et al., 155 2013). The upper 11 cm were dated using a constant rate of supply (CRS) 210Pb age model 156 (Appleby and Oldfield, 1983) (Table 2). The resulting calibrated dates were used in the BACON 157 code which uses Markov chain Monte Carlo statistics to create age-depth models and uses 158 posterior probabilities to determine radiocarbon outliers (Blaauw and Christen, 2011) in the 159 statistical software package “R”(R Core Development Team, 2008). 160 Page 8 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 8 161 Elemental Analysis 162 Half centimeter thick slices were sampled at 1 to 3 cm intervals down the entire length of 163 the core with higher resolution sampling in the upper sediment in order to validate the 164 geochemical record against the historical record. Twenty-one replicate sediment samples were 165 extracted from overlapping cores from a depth of 60.5 to 120.5 cm. Results were generally 166 within 3 µg/g for Pb, 0.2 µg/g for As, 0.006 µg/g for Sb, and 0.004 µg/g for Bi. All samples 167 were lyophilized and homogenized. Elements were extracted using 6 mL of 1 M HNO3 168 overnight, a standard method for extracting weakly bound trace metals from sediments (Graney 169 et al., 1995). Previous research has found that trace metals derived from atmospheric fallout are 170 most frequently present as particulates, which are most commonly weakly adsorbed to clay 171 surfaces and organic matter (Hilton et al., 1985). The supernatant was extracted and diluted 172 before being measured on a Perkin Elmer NeXION 300X inductively coupled plasma mass 173 spectrometer at the University of Pittsburgh. Duplicates were run every 20 samples and were 174 generally within 10% of each other. Blanks were run every 20 samples to check for memory 175 effects and bleed through was consistently below the detection limits of the instrument. 176 177 Results 178 Radiocarbon ages indicate continuous, relatively linear sedimentation from ~16,000 cal 179 years BP to present-day (Table 1) but in this paper, we focus on the last 5000-4000 years of the 180 record (Fig. 2C) as the geochemical variations in this period encompass the most relevant 181 archaeological and historical events. From 3000 to 1200 BC, sediments are composed of 182 unhumified organic-rich sediment, which contains abundant discrete plant macrofossils such as 183 Page 9 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 9 moss (Fig. 2A). From 1200 BC to present-day, sediments gradually transition to homogenous 184 humified organic-rich sediment, characterized by a lack of abundant macrofossils. Organic 185 matter content is relatively high and ranges between 30 and 45% (Fig. 2B). No carbonate 186 material is present throughout the core. 187 Prior to 400 300 BC, concentrations of lead (Pb) and bismuth (Bi) are remarkably stable, 188 averaging 11.19±1.56 µg/g for Pb and 0.01±0.003 µg/g for Bi (Fig. 3). Antimony (Sb) and 189 arsenic (As) are more variable, averaging 0.007±0.006 µg/g for Sb and 0.92±0.13 µg/g for As. 190 The exception to this is a single sample where Sb concentrations increase to 0.05 µg/g at 560 510 191 BC. Henceforth, we will refer to the time prior to 400 300 BC as the “background period” given 192 the relative stability of the geochemistry. Beginning at 400 300 BC metal concentrations begin 193 to increase. They increase more substantially after 210 100 BC and peak at 170 15 BC, before 194 declining to background values by 20 120 AD. Concentrations at 170 15 BC are 7x higher for 195 Pb, 2x higher for Sb, 9x higher for Bi, and 1.5x higher for As than background values. 196 Lead concentrations remain stable up until 1840 1860 AD, but metals such as Sb and Bi 197 show more variability beginning around 1500 AD and generally range between 2 and 4 times 198 that of background values. Arsenic concentrations are also more variable, gradually increasing 199 and declining from 800 to 1100with a peak at 460 AD and then fluctuating between 0.9 and 1.4 200 µg/g from 1250 1000 AD to present. By 1840 1860 AD, concentrations of Pb, Sb, and Bi 201 markedly increase. Lead reaches a peak in 1980 1990 AD of 39.0 µg/g and declines to 28.9 µg/g 202 in the uppermost surface sediments of the core (Fig. 4). In contrast, Bi and Sb concentrations 203 continue to remain high up to the present-day at 0.075 µg/g and 0.15 µg/g, respectively. Arsenic 204 concentrations do notonly show an equivalent increase in the last 200 10 years. 205 206 Page 10 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 10 Discussion 207 Natural airborne sources of Pb, Sb, Bi, and As include volcanoes, soil-derived dust, 208 biogenic processes, and sea salt (Nriagu, 1989; Ferrari et al., 2000); however, anthropogenic 209 sources now comprise the majority of emissions (Nriagu, 1989). The small changes in Pb, Sb, 210 and Bi during the background period (pre-400 300 BC) are likely due to these natural processes, 211 though variability is generally quite low. Known global-scale volcanic events, such as the a 212 large eruption from an unattributed location in 426 BC (Sigl et al., 2015), are not detectable in 213 the Roya sediment core, suggesting that only variations in the local air-shed are responsible for 214 the small background variability at Roya. The single point where Sb is high (0.05 µg/g at 560 215 510 BC) is possibly an anomalous data point given that it does not appear to be part of a larger 216 trend in the dataset. Alternatively, it may reflect a natural source of antimony not derived from 217 mining or human activity. Arsenic displays more variability during the background period which 218 may be because in contrast to the other elements, it can be influenced by lake water pH, 219 oxidation/reduction potential, and uptake by algae (Hamilton-Taylor and Davison, 1978). While 220 the pH of Roya is neutral (Aldasoro et al., 1996), diagenetic effects do have the potential to 221 affect the migration and diffusion of antimony and bismuth, although we suggest that this is not 222 substantial, since trends in antimony and bismuth correlate well with trends in lead (0.58 and 223 0.82, p<0.01, respectively). 224 Notable differences between this study and the previous study by Allen et al., (1996) are 225 the age model. Equivalent depths in the Allen et al., (1996) study are ~1000-2000 years 226 younger, but this studyour work has more robust age control with more than twice as many 227 radiocarbon dates. Allen et al., (1996) found an increased proportion of charcoal, as well as the 228 presence of pollen from cultivated plants, indicating human-induced landscape and vegetational 229 Page 11 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 11 changes at 184 cm (750 BC) (Allen et al., 1996), which our age model would suggest is closer to 230 an age of 2100 BC. Nonetheless, we do not observe any changes in our own geochemical study 231 record at either 750 or 2100 BC, suggesting that human occupation and subsequent landscape 232 change of the Roya watershed is not a substantial driver of the geochemical signal. 233 234 The Roman Period 235 The most prominent feature of the Roya metal record is the two phase increase in metal 236 concentration: phase 1 (400 300 to 210 110 BC) and phase 2 (210 110 BC to 20 120 AD). This 237 increase is very close to the peak in Pb concentrations (Hong et al., 1994) and decline in 238 206Pb/207Pb isotopes from the Greenland ice core (Rosman et al., 1997) that occurs from ~275 to 239 120 BC (Fig. 3), though sampling resolution in the aforementioned studies is too coarse to 240 pinpoint the exact timing of the change. These changes in the Greenland ice core were attributed 241 to Roman lead smelting and are generally in agreement with numerous other geochemical 242 records from Europe (Brannvall et al., 1997; Brannvall et al., 2001; Martínez-Cortizas et al., 243 2013; Renberg et al., 2001) that show widespread lead pollution during the Roman 244 Republic/Empire. 245 We attribute the increases in metals at Roya, particularly Pb, to atmospheric processes. 246 There are no surficial inflows to the lake and its catchment is relatively small. Additionally, the 247 predominant bedrock in the catchment is gneiss (Castro et al., 2003), which is relatively resistant 248 to erosion, and has led to the development of very thin soils (The European Soil Database, 249 2004Askoy et al., 2010) and low angled slopes. All these factors diminish the role that erosional 250 processes play in influencing the geochemical signal. Moreover, other indicators of sediment 251 flux that might explain the peak, such as changes in sedimentation rate (Fig. 2D), or increases in 252 Page 12 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 12 lithogenic metals (Fig. 5) are relatively stable through this period. Lead and other metals are 253 commonly weakly sorbed to organic matter (Renberg, 1986), however weight percent organic 254 matter estimated from LOI at 550ºC does not appreciably change in association with the peak 255 either (Fig. 2B). We believe that it is more likely that atmospheric processes are primarily 256 responsible for metal deposition at this site Ggiven the extremely high concentrations (7x that of 257 background values for Pb), the abrupt increase during phase 2 that takes place over <100 years, 258 the close correspondence to documented metallurgical activity in the region, and the unique 259 setting of the lake, it is more than likely that atmospheric processes are primarily responsible for 260 this metal deposition. The prevailing wind direction is from the southwest while in winter the 261 winds are predominantly from the northwest. The close proximity of Roya to Las Médulas (<35 262 km N) does not preclude the possibility of local, mesoscale atmospheric circulation being 263 responsible for the deposition. Additionally, several smaller Roman mining mine sites occur 264 nearby (El Teleno, Corporales, and Pozos). 265 While the Las Médulas alluvial gold deposits are the closest mining operation to Laguna 266 Roya, approximately 70% of lead deposition in the Greenland ice core during Roman times was 267 attributed to Río Tinto (Rosman et al., 1997) in Southern Spain which was extensively mined for 268 copper and silver (Tylecote, 1992). We do not know the potential transport range of metals to 269 Roya and it is possible that atmospheric transport occurred over ~550 km from Río Tinto, but the 270 concurrent rise in Sb, Bi, and As in the Roya sediment cores (Fig. 3) suggests it is related to is 271 suggestive of the extraction and cupellation of gold. Following the extraction of alluvial gold 272 deposits via hydraulic force, disturbed areas may havelikely dried out, creating dust byproducts 273 that were carried by the wind and deposited at Laguna Roya. However, the extraction of gold 274 via fire-setting was an intensive process in this region as well and temperatures from these fires 275 Page 13 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 13 reached as high as 600-700°C (Weisgerber and Willies, 2000), likely producing a great deal of 276 dust and smoke. 277 Hard-rock gold ores in this region are typically hosted in quartz veins that are rich in 278 arsenopyrite (FeAsS), pyrite (FeS2), galena (PbS), sphalerite ((ZnFe)S), and chalcopyrite 279 (CuFeS2) (Pérez-García et al., 2000; Gómez-Fernández et al., 2012) and further to the south in 280 northern Portugal, bismuthinite (BiS3) is also found (Noronha et al., 2000). Roman slag piles 281 found elsewhere in Europe have high percentages of arsenic and antimony, indications that the 282 Romans extracted trace amounts of gold from minerals such as arsenopyrite and stibnite (Sb2S3) 283 (Healy, 1978). Therefore, we suggest that the increase in metals at Roya is due to local 284 extraction and processing of gold ores given 1) the prevalence of gold veins within solid rock in 285 northwestern Iberia (Fig. 1B), 2) the historical documentation of the widespread use of fire-286 setting in this region (Weisgerber and Willies, 2000), and 3) the geochemical signature in the 287 Roya sediments that matches reasonably well with local ore geology. 288 These results contrast with the Lake Redo study which found a peak in Pb from ~500 to 289 700 AD (Camarero et al., 1998) (Fig. 56) suggesting that the main depositional period was post-290 Roman –Medieval. The Lake Redo study was carried out nearly 18 years ago and relied on bulk 291 sediment dating techniques. Bulk sediments may appear anomalously old due to the 292 incorporation of 14C depleted carbon from carbonate rocks in the watershed (Deevey et al., 293 1954). Such effects were noted and ages were corrected assuming a constant reservoir effect. 294 However, reservoir effects may change through time due to lake level fluctuations and may vary 295 with sediment type and composition (Geyh et al., 1998). Additionally, Redo Lake is surrounded 296 by steep slopes, which are likely prone to rock slides and other material influx. Therefore, it is 297 unclear if the peak in lead is due to atmospheric deposition or terrestrial sediment flux. The 298 Page 14 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 14 timing discrepancies between Redo and La Roya may also point to regional variability in the 299 Iberian Peninsula and a strong role of local mining in atmospheric deposition. 300 However, our results from Roya are broadly consistent with the timing of documented 301 Roman lead deposition in other regional lake and peat bog studies, albeit with a slightly earlier 302 peak increase by a few hundred years. Most studies find the peak of lead concentrations 303 somewhere between 100 BC and 200 AD. The peat bog studies that are closest to our own study 304 site, particularly Penido Vello, observe Pb and other heavy metal deposition associated with 305 Roman activities from ~100 to 200 AD (Martínez-Cortizas et al., 1997) (Fig. 56). However, 306 there are many important regional variations that are observed throughout Europe, reflecting 307 unique histories. Given the proximity of many other known Roman mines to Penido Vello and 308 Molina (Fig. 1), spatial and temporal shifts in mining focus may be responsible for these 309 differing geochemical increases. 310 It is unclear when Las Médulas was first exploited by the Romans. Archaeological finds 311 suggest large-scale exploitation during the early imperial period (~25 BC to 197 AD) (Lewis and 312 Jones, 1970), but earlier use cannot be excluded due to the destructive nature of mining 313 techniques (Edmonson, 1989). Therefore, the results of our study suggest that an earlier, small-314 scale, exploitation of Las Médulas beginning ~400 300 BC is possible. Interestingly, the study 315 that most closely matches our own in terms of timing is from Zoñar Lake in southern Spain 316 where pre-industrial Pb concentrations were highest from 500 to 100 BC (Martín-Puertas et al., 317 2010) (Fig. 56). In this study, the authors suggested that concentrations increased prior to the 318 peak of the Roman Empire due to extensive mining on the the southern and eastern areas of the 319 Iberian Peninsula from Iberian societies influenced Greek and Phoenician societiescolonies. The 320 increases in metals during phase 1 at Roya may partially be explained by Carthaginian mining 321 Page 15 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 15 activities, but the peak of concentrations during phase 2 are too late to be attributed to any 322 civilization but the Romans. 323 Lead is a pollutant that may remain persistent in both the terrestrial (Beyer et al., 1998) 324 and aquatic (Hillman et al., 2015) environment. However, the Roya sediment Pb concentrations 325 return relatively rapidly to within a few µg/g of background values. Heavy metals deposited on 326 the landscape proximal to Laguna Roya are unlikely to be remobilized due to the resistant gneiss 327 bedrock and the low potential for alluvial and colluvial storage locations. The spatial extent of 328 this Pb deposition is difficult to quantify since only two other studies have been conducted in this 329 region. However, the increased Pb deposition in northwestern Spain as a result of Roman 330 activities appears to be clear; within this region, landscapes with greater potential for sediment 331 remobilization may face contemporary contamination issues as a result of the remobilization of 332 this Pb pollution. 333 334 The Medieval and Industrial Periods 335 We do not observe any geochemical variations during the early medieval period that 336 would suggest protracted post-Roman mining or metallurgical activity, contrary to what was 337 inferred from other Spanish lake sediment studies (Camarero et al., 1998; Martínez-Cortizas et 338 al., 2013). The spatial and temporal extent of post-Roman mineral resource exploitation is 339 largely unquantified because the re-opening of abandoned mines within the last 50 years has 340 destroyed previously occupied sites and focused only on hard-rock mines as opposed to alluvial 341 deposits (Edmonson, 1989), such as Las Médulas. However, the Roya sediment record does not 342 show evidence of regional mineral extraction and processing after ~210 120 AD. 343 Page 16 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 16 The increases in both Sb and Bi that take place between 1500 and 1800 1700 AD may be 344 the result of mining and smelting associated with the rise of the Spanish Empire. Neither Pb nor 345 As show a contemporaneous increase, suggesting that a metallurgical process different than the 346 Roman mode of exploitation is responsible for these increases. A study of several Swedish lakes 347 found Pb concentration and isotopic changes beginning around 1400 AD that were attributed to 348 new techniques of silver extraction from copper ores that required large amounts of lead 349 (Brannvall et al., 2001). However, the increases in As and Bi at Roya occur without concomitant 350 increases in Pb. It is difficult to attribute these increases to any particular process because little 351 is known about late medieval Spanish metallurgy aside from the increasing skill and production 352 of ferrous metal working in northern Spain (Tylecote, 1992). However, concentrations of Ni, 353 Zn, As, and Cd at Molina Mire increase beginning around 1300 AD (Pontevedra-Pombal et al., 354 2013), which points to a possible coherent period of regional mineral resource exploitation in 355 northwestern Spain during this time period. The lowest Sb, Bi and As concentrations occur 356 around 1800 AD, a period of political and economic crises in Spain (Barreiro-Lostres et al., 357 2015). Further investigation at Roya and other regional lakes would be needed to more 358 definitively characterize the impacts of medieval and post-medieval metallurgy and this is an 359 area of potential future study for this region. 360 The changes in Pb concentrations at Roya over the past several hundred years closely 361 correspond to known historical events (Fig. 4). Concentrations begin to increase around the start 362 of the Industrial Revolution and may be attributed to a variety of large-scale industrial processes 363 that utilized lead as a flux material (Tylecote, 1992). Concentrations continue to increase with 364 the advent of leaded gasoline in the 1920’s, which resulted in the widespread atmospheric 365 deposition of lead throughout Europe that peaked around 1970 AD (Renberg et al., 2001). 366 Page 17 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 17 Concentrations at Roya peak a bit later in 1980 1990 AD (Fig. 4). Many countries in Europe 367 began phasing out leaded gasoline in the 1970’s and 1980’s, which agrees well with our results 368 from Roya, although Spain was given additional time to comply with such regulations, possibly 369 explaining the slightly later peak in lead at Roya. The close correspondence between Pb 370 concentrations and historical documentation further supports our supposition that Pb 371 concentrations can be used to infer past anthropogenic changes in the metal cycle that occurred 372 proximal to the lake. While present-day concentrations of As have remained relatively low 373 compared to the historic period, concentrations of Sb and Bi are equal to or exceed the peak of 374 Roman pollution. These modern-day increases are likely the result of industrial emissions which 375 are estimated to be larger than natural emissions by as much 100-200% (Nriagu, 1989). 376 In this paper, we have documented the contemporaneous environmental consequences of 377 Roman mining activities taking place near the extensive Las Médulas gold mine. We find 378 evidence for mineral resource exploitation beginning at ~400 300 BC, slightly earlier than was 379 previously supposed proposed for Roman activities in this region (Lewis and Jones, 1970). This 380 regional metal pollution continued for 610 420 years and left a profound geochemical signature 381 in the Laguna Roya sediments. The concentrations of Pb at Roya at the peak of Roman 382 exploitation are double that of the last hundred years. These findings are somewhat unique as 383 only a handful of studies have found pre-industrial pollution to be greater in magnitude than that 384 of modern-day (Abbott and Wolfe, 2003; Hillman et al., 2015), including the study of Redo Lake 385 in the Pyrenees (Camarero et al., 1998). This study adds to our understanding of large-scale 386 Roman mineral resource exploitation and finds that the produced environmental consequences at 387 this lake can beare substantial and exceed or rival that of today’s lead pollutionenvironmental 388 degradation. 389 Page 18 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 18 390 ACKNOWLEDGEMENTS: 391 We thank Matthew Finkenbinder and Ashley Albert for help with laboratory procedures 392 and Jordan Abbott for assistance with fieldwork. We thank two anonymous reviewers for their 393 helpful comments and suggestions. Aubrey Hillman recognizes the support of the Byrd Post-394 Doctoral Research Fellowship during the preparation of this manuscript. We thank the financial 395 support of the Hewlett Foundation, the University of Pittsburgh International Studies Fund, the 396 CSIC I-Link program (I-LINK0510), the MINECO Ministry of Spain (MEDLANT project 397 CGL2016-76215-R), and the US National Science Foundation (EAR/IF grant 0948366). 398 399 REFERENCES: 400 Abbott MB and Stafford TW. (1996) Radiocarbon Geochemistry of Modern and Ancient Arctic 401 Lake Systems, Baffin Island, Canada. Quaternary Research 45: 300-311. 402 Abbott MB and Wolfe AP. (2003) Intensive pre-Incan metallurgy recorded by lake sediments 403 from the Bolivian Andes. Science 301: 1893-1895. 404 Aldasoro JJ, Aedo C, Munoz J, et al. (1996) A Survey of Cantabrian Mires (Spain). Anales del 405 Jardín Botánico de Madrid 54: 472-489. 406 Allen JRM, Huntley B and Watts WA. (1996) The vegetation and climate of northwest Iberia 407 over the last 14 000 yr. Journal of Quaternary Science 11: 125-147. 408 Appleby P and Oldfield F. (1983) The assessment of 210Pb data from sites with varying 409 sediment accumulation rates. Hydrobiologia 103: 29-35. 410 Askoy E, Panagos P, Montanarella L, et al. (2010) European Soil Database. In: Bureau ES (ed). 411 Page 19 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 19 Bacardit M, Krachler M and Camarero L. (2012) Whole-catchment inventories of trace metals in 412 soils and sediments in mountain lake catchments in the Centrol Pyrenees: Apportioning 413 the anthropogenic and natural contributions. Geochemica et Cosmochimica Acta 82: 52-414 67. 415 Barreiro-Lostres F, Brown E, Moreno A, et al. (2015) Sediment delivery and lake dynamics in a 416 Mediterranean mountain watershed: Human-climate interactions during the last 417 millennium (El Tobar Lake record, Iberian Range, Spain). Sci Total Environ 533: 506-418 519. 419 Beyer WN, Auget DJ, Morton A, et al. (1998) Lead Exposure of Waterfowl Ingesting Coeur 420 d'Alene River Basin Sediments. J Environ Qual 27: 1533-1538. 421 Bindler R, Yu R, Hansson M, et al. (2012) Mining, Metallurgy and the Historical Origin of 422 Mercury Pollution in Lakes and Watercourses in Central Sweden. Environmental Science 423 and Technology 46: 7984-7991. 424 Blaauw M and Christen JA. (2011) Flexible Paleoclimate Age-Depth Models Using an 425 Autoregressive Gamma Process. Bayesian Analysis 6: 457-474. 426 Brannvall M-L, Bindler R, Emteryd O, et al. (1997) Stable lead isotopes and concentration 427 records of atmospheric lead pollution in peat and lake sediments in Sweden. Water Air 428 Soil Pollution 100: 243-252. 429 Brannvall M-L, Bindler R, Emteryd O, et al. (2001) Four thousand years of atmospheric lead 430 pollution in northern Europe: a summary from Swedish lake sediments. Journal of 431 Paleolimnology 25: 421-435. 432 Page 20 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 20 Camarero L, Masque P, Devos W, et al. (1998) Historical Variations in Lead Fluxes in the 433 Pyrenees (Northeast Spain) from a Dated Lake Sediment Core. Water, Air, and Soil 434 Pollution 105: 436-449. 435 Castro A, Corretge LG, De La Rosa JD, et al. (2003) The Appinite-Migmatite Complex of 436 Sanabria, NW Iberian Massif, Spain. Journal of Petrology 44: 1309-1344. 437 Center UWH. (2015) 438 Cooke CA, Abbott M, Wolfe AP, et al. (2007) A Millennium of Metallurgy Recorded by Lake 439 Sediments from Morococha, Peruvian Andes. Environmental Science and Technology 41: 440 3469-3474. 441 Dean WE. (1974) Determination of carbonate and organic matter in calcareous sediments and 442 sedimentary rocks by loss on ignition: comparison with other methods. Journal of 443 Sedimentary Petrology 44: 242-248. 444 Deevey ES, Gross MS, Hutchinson G, et al. (1954) The natural C14 contents of materials from 445 hard-water lakes. Geology 40: 285-288. 446 Edmonson JC. (1989) Mining in the Later Roman Empire and beyond: Continuity or Disruption? 447 The Journal of Roman Studies 79: 84-102. 448 Ferrari CP, Hong S, Van de Velde K, et al. (2000) Natural and anthropogenic bismuth in Central 449 Greenland. Atmospheric Environment 34: 941-948. 450 Gallon C, Tessier A, Gobeil C, et al. (2004) Modeling diagenesis of lead in sediments of a 451 Canadian Shield lake. Geochimica et Cosmochimica Acta 68: 3531-3545. 452 García-Alix A, Jimenez-Espejo FJ, Lozano JA, et al. (2013) Anthropogenic impact and lead 453 pollution throughout the Holocene in Southern Iberia. Science of The Total Environment 454 449: 451-460. 455 Page 21 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 21 Geyh M, Schotterer U and Grosjean M. (1998) Temporal changes of the 14C reservoir effect in 456 lakes. Radiocarbon 40: 921-931. 457 Gómez-Fernández F, Vindel E, Martín-Crespo T, et al. (2012) The Llamas de Cabrera gold 458 district, a new discovery in the Variscan basement of northwest Spain: A fluid inclusion 459 and stable isotope study. Ore Geology Reviews 46: 68-82. 460 Graney J, Halliday A, Keeler G, et al. (1995) Isotopic record of lead pollution in lake sediments 461 from the northeastern United States. Geochemica et Cosmochimica Acta 59: 1715-1728. 462 Hamilton-Taylor J and Davison W. (1978) Redox-Driven Cycling of Trace Elements in Lakes. 463 In: Lerman A, Imboden DM and Gat JR (eds) Physics and Chemistry of Lakes. 2nd ed. 464 New York: Springer-Verlag. 465 Healy JF. (1978) Mining and Metallurgy in the Greek and Roman World, London: Thames and 466 Hudson. 467 Hillman AL, Abbott MB, Yu JQ, et al. (2015) The Environmental Legacy of Copper Metallurgy 468 and Mongol Silver Smelting Recorded in Yunnan Lake Sediments. Environmental 469 Science and Technology 49: 3349-3357. 470 Hilton J, Davison W, Ochsenbein U. (1985) A Mathematical Model for Analysis of Sediment 471 Core Data: Implications for Enrichment Factor Calculations and Trace-Metal Transport 472 Mechanisms. Chemical Geology 48: 281-291. 473 Hong S, Candelone J, Patterson CC, et al. (1994) Greenland Ice Evidence of Hemispheric Lead 474 Pollution Two Millennia Ago by Greek and Roman Civilizations. Science 265: 1841-475 1843. 476 Page 22 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 22 Global Network of Isotopes in Precipitation,. The GNIP Database. IAEA/WMO, 22 Dec. 2016. 477 (2014) Global Network of Isotopes in Precipitation. The GNIP Database. Available at: 478 http://www.iaea.org/water. 479 Kylander M, Weiss D, Martínez-cCortizas A, et al. (2005) Refining the pre-industrial 480 atmospheric Pb isotope evolution curve in Europe using an 8000 year old peat core from 481 NW Spain. Earth and Planetary Science Letters 240: 467-485. 482 Leblanc M, Morales JA, Borrego J, et al. (2000) 4,500-year-old mining pollution in 483 Southwestern Spain: long-term implications for modern mining pollution. Economic 484 Geology 95: 655-662. 485 Lee CSL, Qi S-H, Zhang G, et al. (2008) Seven Thousand Years of Records on the Mining and 486 Utilization of Metals from Lake Sediments in Central China. Environmental Science and 487 Technology 42: 4732–4738. 488 Lewis PR and Jones GDB. (1970) Roman Gold-Mining in North-West Spain. The Journal of 489 Roman Studies 60: 169-185. 490 López-Merino L, Martínez-Cortizas A, Reher GS, et al. (2014) Reconstructing the impact of 491 human activities in a NW Iberian Roman mining landscape for the last 2500 years. 492 Journal of Archaeological Science 50: 208-218. 493 Martín-Puertas C, Jiménez-Espejo F, Martínez-Ruiz F, et al. (2010) Late Holocene climate 494 variability in the southwestern Mediterranean region: an integrated marine and terrestrial 495 geochemical approach. Climate of the Past 6: 807-816. 496 Martínez-Cortizas A, Pontevedra-Pombal X, Novoa Munoz JC, et al. (1997) Four Thousand 497 Years of Atmospheric Pb, Cd, and Zn Deposition Recorded by the Ombrotrophic Peat 498 Bog of Penido Vello (Northwestern Spain). Water, Air, and Soil Pollution 100: 387-403. 499 Page 23 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 23 Martínez-Cortizas A, López-Merino L, Bindler R, et al. (2013) Atmospheric Pb pollution in N 500 Iberia during the late Iron Age/Roman times reconstructed using the high-resolution 501 record of La Molina mire (Asturias, Spain). Journal of Paleolimnology 50: 71-86. 502 Mighall TM, Timberlake S, Foster IDL, et al. (2009) Ancient copper and lead pollution records 503 from a raised bog complex in Central Wales, UK. Journal of Archaeological Science 36: 504 1504-1515. 505 Muñoz Sobrino C, Heiri O, Hazekamp M, et al. (2013) New data on the Lateglacial period of 506 SW Europe: a high resolution multiproxy record from Laguna de la Roya (NW Iberia). 507 Quaternary Science Reviews 80: 58-77. 508 Noronha F, Cathelineau M, Boiron M-C, et al. (2000) A three stage fluid flow model for 509 Variscan gold metallogenesis in northern Portugal. Journal of Geochemical Exploration 510 71: 209-224. 511 Nriagu J. (1989) A global assessment of natural sources of atmospheric trace metals. Nature 338: 512 47-49. 513 Pérez-García LC, Sanchez-Palencia FJ and Torres-Ruiz J. (2000) Tertiary and Quaternary 514 alluvial gold deposits of Northwest Spain and Roman mining (NW of Duero and Bierzo 515 Basins). Journal of Geochemical Exploration 71: 225-240. 516 Pliny. (1952) Natural History. Cambridge: MA: Harvard University Press. 517 Pompeani DP, Abbott M, Steinman B, et al. (2013) Lake Sediments Record Prehistoric Lead 518 Pollution Related to Early Copper Production in North America. Environmental Science 519 and Technology 47: 5545-5552. 520 Page 24 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 24 Pontevedra-Pombal X, Mighall TM, Nóvoa-Muñoz JC, et al. (2013) Five thousand years of 521 atmospheric Ni, Zn, As, and Cd deposition recorded in bogs from NW Iberia: prehistoric 522 and historic anthropogenic contributions. Journal of Archaeological Science 40: 764-777. 523 R Core Development Team. (2008) R: A language and environment for statistical computing. 524 Reimer PJ, Bard E, Bayliss A, et al. (2013) INTCAL13 and Marine 13 Radiocarbon Age 525 Calibration Curves 0–50,000 years CAL BP. Radiocarbon 55: 1869-1887. 526 Renberg I. (1986) Concentration and annual accumulation values of heavy metals in lake 527 sediments: Their significance in studies of the history of heavy metal pollution. 528 Hydrobiologia 143: 379-385. 529 Renberg I, Bindler R and Brannvall M-L. (2001) Using the historical lead-deposition record as a 530 chronological marker in sediment deposits in Europe. The Holocene 11: 511-516. 531 Richardson JS. (1976) The Spanish Mines and the Development of Provincial Taxation in the 532 Second Century B.C. The Journal of Roman Studies 66: 139-152. 533 Rosman KJR, Chisholm W, Hong S, et al. (1997) Lead from carthainian Carthainian and roman 534 Roman Spanish mines isotopically identified in Greenland ice dated from 600 B.C. to 535 300 A.D. Environmental Science and Technology 31: 3413-3416. 536 Ruiz del Árbol Moro M, Sánchez-Palencia Ramos FJ, Sastre Prats I, et al. (2014) Water 537 networks of Roman gold mines of Northwestern Iberian Peninsula. Water History 6: 95-538 113. 539 Shotyk W, Weiss D, Appleby P, et al. (1998) History of Atmospheric Lead Deposition since 540 12,370 14C yr BP from a Peat Bog, Jura Mountains, Switzerland. Science 281: 1635-541 1640. 542 Page 25 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 25 Sigl M, Winstrup M, McConnell JR, et al. (2015) Timing and climate forcing of volcanic 543 eruptions for the past 2,500 years. Nature 523: 543-549. 544 R Core Development Team. (2008) R: A language and environment for statistical computing. 545 The European Soil Database distribution version 2.0, European Commission and the European 546 Soil Bureau Network, CD-ROM, EUR 19945 EN, 2004. 547 Tylecote RF. (1992) A History of Metallurgy, London: Maney Publishing. 548 UNESCO World Heritage Center, Las Médulas, 22 Dec. 2016. Available at: 549 http://whc.unesco.org/en/list/803/. 550 Weisgerber G and Willies L. (2000) The Use of Fire in Prehistoric and Ancient Mining : 551 Firesetting. Paléorient 26: 131-149. 552 Wright H, Mann D and Glaser P. (1984) Piston corers for peat and lake sediments. Ecology 65: 553 657-659. 554 Page 26 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review Fig. 1. A) 1- Laguna Roya, 2- Penido Vello, 3- Molina Mire, 4- Zoñar Lake, 5- Redo Lake; B) Known Roman alluvial gold mines (red circles) and gold veins in solid rock (purple squares) adapted from Pérez-García et al., 2000; C) Watershed of Laguna Roya with surficial geology adapted from Castro et al., 2003. Fig. 1 139x238mm (600 x 600 DPI) Page 27 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review Fig. 2. A) Stratigraphic column; B) Weight percent organic matter estimated from LOI 550°C; C) Age-depth model with 95% confidence intervals and radiocarbon dates (blue circles) and 210Pb dates (green triangles) with 2-sigma error bars; D) Sedimentation rate in cm/year. Fig. 2 116x87mm (600 x 600 DPI) Page 28 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review Fig. 3. A) Archaeological and cultural periods of the Mediterranean region; B) concentrations of lead from the Greenland ice core (Hong et al., 1994); C) 206Pb/207Pb ratio from the Greenland ice core (Rosman et al., 1997), note the flipped axis; D) Roya lead (Pb) concentrations; E) antimony (Sb); F) bismuth (Bi); G) arsenic (As). The peak of metal concentrations at Roya is close in timing to that of the Greenland ice core. Fig. 3 186x271mm (600 x 600 DPI) Page 29 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review Fig. 4. Lead concentrations over the last ~310 years showing a close correspondence between increases in concentration and historical events. Fig. 4 77x28mm (600 x 600 DPI) Page 30 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review Fig. 5. Comparison of lead concetrations (A) at Roya with those of lithogenic elements (B- potassium, C- magnesium, D- aluminum) over the last 4,000 years. In general, the increases in lead concentration are not closely correlated to increases in lithogenic elements. Fig. 5 200x217mm (600 x 600 DPI) Page 31 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review Fig. 6. Comparison of lead pollution records from Spain- A) Laguna Roya lead concentrations; B) Penido Vello lead enrichment factor (Martínez-Cortizas et al., 1997); C) Molina mire lead enrichment factor (Martínez-Cortizas et al., 2013); D) Redo Lake lead concentrations (Camarero et al., 1998); E) Zoñar Lake lead to aluminum ratio (Martín-Puertas et al., 2010). Fig. 6 154x222mm (600 x 600 DPI) Page 32 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review Table 1. AMS radiocarbon dates. UCI Number Composite Core Depth (cm) Material 14 C age (BP) Error ± Median Probability Calibrated Age (yr BP) 2σ Calibrated Age Range (yr BP) 180956 26 Charcoal 725 15 675 665-685 164748 46 Wood 1125 30 1023 959-1172 164749 62 Charcoal 1460 30 1346 1302-1396 152027 89.5 Charcoal 2295 15 2336 2316-2349 180957 116.5 Charcoal 2875 45 2979 2874-3084 180958 132.5 Charcoal 3150 70 3359 3207-3512 152028 148.5 Moss 3470 15 3753 3651-3828 152029 209.5 Moss 3940 15 4415 4298-4438 152030 235.5 Moss 4260 20 4842 4829-4856 164750 336.5 Moss 5250 30 5996 5926-6177 164751 385.5 Wood 6460 250 7334 6756-7821 152031 427 Charcoal 7250 20 8062 8007-8159 152032 506.5 Leaf 8980 30 10186 9941-10232 152033 552 Leaf 9925 30 11307 11244-11398 152034 587.5 Moss 10350 30 12207 12034-12388 152035 638.5 Moss 12945 40 15468 15275-15680 Page 33 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review Table 2. Down-core 210 Pb activities, 214 Pb activities, cumulative weight, and CRS sediment ages. Depth (cm) 210 Pb activity, Bq-g -1 1σ Error 210 Pb activity 214 Pb activity, Bq-g -1 1σ Error 214 Pb activity Cumulative Weight, g cm -1 CRS age (yr AD/BC) 1σ Error Age Cores A-09 0.0-0.5 0.5860 0.1615 0.1700 0.0560 0.0158 2014 1.16 1.0-1.5 0.8690 0.0970 0.0657 0.0146 0.0689 2010 1.23 2.0-2.5 0.8290 0.0905 0.0000 0.0000 0.1342 2005 1.35 3.0-3.5 0.8540 0.1005 0.0000 0.0000 0.1751 2001 1.42 4.0-4.5 0.8310 0.0895 0.0356 0.0080 0.2509 1994 1.67 5.0-5.5 0.7630 0.0815 0.0448 0.0090 0.3418 1983 2.06 6.0-6.5 0.5250 0.0580 0.0436 0.0069 0.4388 1972 2.37 7.0-7.5 0.4200 0.0487 0.0326 0.0075 0.5401 1958 2.98 8.0-8.5 0.3270 0.0406 0.0412 0.0069 0.6498 1941 4.15 9.0-9.5 0.2570 0.0345 0.0529 0.0077 0.7400 1923 5.50 10.0-10.5 0.2280 0.0318 0.0511 0.0079 0.8207 1896 10.63 11.0-11.5 0.1610 0.0296 0.0647 0.0076 0.8960 1843 46.66 Page 34 of 33 http://mc.manuscriptcentral.com/holocene HOLOCENE 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60