Science of the Total Environment 470–471 (2014) 707–716

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Variation and distribution of and metalloids in soil/ash mixtures from Agbogbloshie e-waste recycling site in ,

Takaaki Itai a,⁎, Masanari Otsuka a, Kwadwo Ansong Asante a,b,MamoruMutoa, Yaw Opoku-Ankomah b,OsmundDuoduAnsa-Asareb, Shinsuke Tanabe a a Center for Marine Environmental Studies, Ehime University, Bunkyo-cho 2-5, Matsuyama, Ehime 790-8577, Japan b CSIR Water Research Institute, P. O. Box AH 38, Achimota, Accra, Ghana

HIGHLIGHTS

• Contamination on the largest e-waste recycling site in Africa was investigated. • Portable X-ray Fluorescence analyzer useful for first screening • High levels of Cu, Zn, Pb, and Al in soil/ash mixtures • for workers are significant.

article info abstract

Article history: Illegal import and improper recycling of (e-waste) are an environmental issue in developing Received 29 April 2013 countries around the world. African countries are no exception to this problem and the Agbogbloshie market Received in revised form 9 October 2013 in Accra, Ghana is a well-known e-waste recycling site. We have studied the levels of (loid)s in the mixtures Accepted 10 October 2013 of residual ash, formed by the burning of e-waste, and the cover soil, obtained using a portable X-ray fluorescence Available online 30 October 2013 spectrometer (P-XRF) coupled with determination of the 1 M HCl-extractable fraction by an inductively coupled Keywords: plasma mass spectrometer. The accuracy and precision of the P-XRF measurements were evaluated by measuring E-waste 18 standard reference materials; this indicated the acceptable but limited quality of this method as a screening Ghana tool. The HCl-extractable levels of Al, Co, Cu, Zn, Cd, In, Sb, Ba, and Pb in 10 soil/ash mixtures varied by Trace element more than one order of magnitude. The levels of these metal(loid)s were found to be correlated with the color Portable XRF (i.e., soil/ash ratio), suggesting that they are being released from disposed e-waste via open burning. The source Risk assessment of rare elements could be constrained using correlation to the predominant metals. Human quotient values based on ingestion of soil/ash mixtures exceeded unity for Pb, As, Sb, and Cu in a high-exposure scenario. This study showed that along with common metals, rare metal(loid)s are also enriched in the e-waste burning site. We suggest that risk assessment considering exposure to multiple metal(loid)s should be addressed in studies of e-waste recycling sites. © 2013 Elsevier B.V. All rights reserved.

1. Introduction recycling can induce severe environmental contamination, as e-wastes contain more than 1000 different chemical substances, many of Rapidly developing countries face problems because of which are toxic, such as , , , , selenium, increasing amounts of electronic waste (e-waste) from both domestic hexavalent chromium, and flame retardants that create dioxin generation and illegal imports. For emerging economies, these material emissions when burned (Buser, 1986; Leung et al., 2011). Because flows from waste imports not only offer a business opportunity but workers in recycling areas are exposed to this complex mixture of also satisfy the demand for cheap secondhand electrical and electronic toxic materials, deciding suitable risk assessment criteria is always equipment (Widmer et al., 2005). Uncontrolled processes of e-waste complicated. Recently, a number of reports have noted the serious contamination in Asian developing countries (Terazono et al., 2006), e.g., in Guiyu, China (Huo et al., 2007; Leung et al., 2006). There are some large-scale recycling sites other than in China, e.g., Delhi and Bangalore in India, ⁎ Corresponding author. Tel.: +81 89 927 8132; fax: +81 89 927 8133. and Accra in Ghana. Our group recently undertook monitoring surveys E-mail address: [email protected] (T. Itai). in various e-waste recycling areas in Bangalore (Ha et al., 2009)and

0048-9697/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.scitotenv.2013.10.037 708 T. Itai et al. / Science of the Total Environment 470–471 (2014) 707–716

(a) (b)

Ghana Africa Accra

Atlantic Ocean

(c)

Fig. 1. Location map of study area. (a) Location of Ghana and Accra in Africa; (b) location of the Agbogbloshie market; (c) enlarged map of the study site. The dotted circle in map (c) indicates the site for burning e-waste. The color in map (c) indicates surface soil color with reference to a Google Earth satellite image in January 2010 (Fig. S1).

Bui Dau, Vietnam (Tue et al., 2010). These studies showed apparent previous reports have suggested that metals and organic contaminants, contamination by metal(loid)s and organic pollutants in these sites. such as brominated flame retardants (BFRs), polychlorinated dibenzo The present study discusses the result of our recent survey in dioxins (PCDD) and furans (PCDF) are associated with e-waste burning Agbogbloshie market, the largest e-waste recycling site in Accra, (Atiemo et al., 2012; Bridgen et al., 2008; Caravanos et al., 2011; Oteng- Ghana. The Republic of Ghana has an unregulated and unrestricted Ababio, 2012). A previous human monitoring survey by our group import regime for secondhand electrical and electronic equipment. suggested that recycling workers in Agbogbloshie have high urinary Any e-waste could enter the country under the guise of secondhand levels of some metal(loid)s, particularly Fe, Sb, and Pb (Asante et al., electrical and electronic equipment uncontrolled. The demand 2012). This situation suggested the necessity for a further monitoring for such equipment in Ghana continues to grow and in 2009, survey and risk assessment. In addition, residues of burned waste these imports totaled 215,000 tons, equivalent to 9 kg per capita possibly contain not only toxic metals but also rare metal(loid)s (Amoyaw-Osei et al., 2011). Most of the e-waste comes from Europe (e.g., In, Sb, Bi). However, reports on these metals are relatively scarce. and North America. At Agbogbloshie market in Accra, heaps of old A characteristic of metal(loid) contamination in e-waste recycling computers and their accessories and other electronic goods that are not sites is its spatially heterogeneous distribution. Generally, extra- usable are dumped continuously with no concern for the hazards that ordinarily high levels of metal(loid)s have been found from residual they may pose to the environment and the people living nearby. Some ash after burning of e-waste, while these levels are usually attenuated

(a) (b)

Fig. 2. The picture of working environment in Agbogbloshie e-waste recycling site. (a) Dismantling of cables, circuit boards, and other parts from PC. (b) Open burning of e-wastes. T. Itai et al. / Science of the Total Environment 470–471 (2014) 707–716 709 with increasing distance from the source (Li et al., 2011). In this study, Considering this background, we set the objectives of this study we hypothesized that the color of soil/ash mixtures could give a good as the following: (i) measuring metal(loid) levels in soil/ash mixtures correlation to the levels of metal(loid)s because the color of a sample using a portable X-ray fluorescence spectrometer (P-XRF) as well as becomes darker as the proportion of ash increases. Because the color the 1 M HCl-extractable fraction; (ii) assessing the geographical of surface soil can be observed with high spatial resolution using distribution of metal(loid)s by focusing on soil color; (iii) constraining satellite technology, it may be a valuable approach for seeing the picture the source of rare elements using correlation to the predominant of metal(loid) distribution in e-waste recycling sites. metals; (iv) comparing metal(loid) levels in soil/ash mixtures with the

20000 2000 100000

80000 15000 1500

60000 10000 1000 40000

5000 500 20000 Ti certified value (mg/kg) Fe certified value (mg/kg) R2=0.602 Mn certified value (mg/kg) R2=0.867 R2=0.989 0 0 0 0 5000 10000 15000 20000 0 500 1000 1500 2000 0 40000 80000 Ti measured by P-XRF (mg/kg) Mn measured by P-XRF (mg/kg) Fe measured by P-XRF (mg/kg)

250 2500 50

200 2000 40

150 1500 30

100 1000 20

50 500 10 As certified value (mg/kg) Zn certified value (mg/kg) Cu certified value (mg/kg) R2=0.942 R2=0.997 R2=0.876 0 0 0 0 50 100 150 200 250 0 500 1000 1500 2000 2500 0 1020304050 Cu measured by P-XRF (mg/kg) Zn measured by P-XRF (mg/kg) As measured by P-XRF (mg/kg)

500 500 7000

6000 400 400 5000 300 300 4000

200 200 3000

2000 100 100 1000 Sr certified value (mg/kg) Ba certified value (mg/kg) Rb certified value (mg/kg) 2 R =0.995 R2=0.983 R2=0.643 0 0 0 0 100 200 300 400 500 0 100 200 300 400 500 0 1000 2000 3000 4000 5000 6000 7000 Rb measured by P-XRF (mg/kg) Sr measured by P-XRF (mg/kg) Ba measured by P-XRF (mg/kg)

500

400

300 R2=0.988 200

100 Pb certified value (mg/kg) 0 0 100 200 300 400 500 Pb measured by P-XRF (mg/kg)

Fig. 3. Relationship of 10 metal(loid) concentrations in 18 certified geological materials between the certified value and that measured by P-XRF. The 1:1 line is given by the dotted line; the solid line represents the approximate curve. 710 T. Itai et al. / Science of the Total Environment 470–471 (2014) 707–716 previous urinary data of Asante et al. (2012); and (v) assessing risk Research (CSIR), Water Research Institute in Accra, then transported according to ingestion of the soil/ash mixture. to Japan by airplane with freezing by gel ice; subsequently, they were registered to the Environmental Specimen Bank (es-Bank) at Ehime 2. Materials and methods University, which has archived many environmental/biological samples over the last 40 years at −25 °C (Tanabe, 2006). Prior to the chemical 2.1. Site description analyses, ca. 5 g of subsamples was air-dried and ground after removal of some components other than soil/ash mixtures, such as grass or The Agbogbloshie market is situated on a bank of the Odaw River metallic wire. and in the upper reaches of the Korle Lagoon (Fig. 1). In this area, Concentrations of 15 elements in soil/ash mixtures were disused computers, TVs, and other electronic and electrical devices as measured using a P-XRF spectrometer with Ag anode X-ray tube well as household waste are dumped indiscriminately (Fig. 2). This operated at 40 keV (Omega Xpress PXRF; Innov-X Systems, Inc., market has become the hub of an informal “recycling” industry in Woburn, MA, USA) after the samples were brought to the laboratory. Ghana. Before the measurements, standardization of the P-XRF was made The study site is a flat and heavily industrialized area that consists of via measurement of stainless steel 316 alloy. We employed “soil scattered recyclers working out of small sheds and in the open (Fig. 1). mode,” which has a scatter-normalization algorithm for soil, The frontage of the site adjacent to the Agbogbloshie market is where allowing for sequential analysis via K-line and L-line electron the dismantling and sale of metal(loid)s are conducted. Open burning fluorescence (US EPA, 2007). One gram of soil/ash mixture was of e-waste occurs at the rear of the market to recover valuable metals placed in a cell with its bottom sealed by a clear film to such as Cu. Most workers work without masks, thereby exposing allow detection of emitted fluorescence X-ray from the sample. themselves to fly ash emitted from the burning sites (Fig. 2). The Samples were scanned for 120s. Elemental concentrations and stan- distance between Agbogbloshie market and burning places in which dard errors were reported for individual elements in each scan. We we have collected soil/ash mixtures was less than 300 m. evaluated the accuracy by duplicate measurements of 18 certified geological materials distributed by the Geological Society of Japan, 2.2. Sample collection and chemical analysis including igneous rocks (JA-3, JB-1b, JB-3, JG-3, JGb-2, JR-1, JR-3, JP-2), sedimentary rock/sediments (JLs-1, JSl-1, JSl-2, JLk-1, JSd-1, JSd-2, JMS-1, Ten soil/ash mixtures, E1 to E10 specified in Fig. 1, were collected JMS-2), coal fly ash (JCFA-1), and an ore sample (JMn-1). The recovery from Agbogbloshie market in August 2010. During our sampling values and certified concentrations of metal(loid)s are given in campaign, we observed that the burning of e-waste occurred mainly Tables S1 and S2. Although the recovery values are concentration in two locations (Fig. 1). The ground color was dark around the burning dependent, common trace metal(loid)s including Fe, Cu, Zn, As, sites and gradually changed to a lighter color with increasing distance Rb, Sr, and Pb gave acceptable quality with the geometric mean by mixing with the cover soil. The soil/ash mixture samples (0–2cm value of recovery being 85–113% (n = 10–16). depth), which have various colors, were collected using a stainless Chemical extraction with 1 M HCl was also carried out to assess steel auger. A composite soil/ash mixture consisting of five subsamples the bioaccessible fraction. The following procedure is the standard within 1m2 was collected at each point. These samples were first kept in method provided by the Japanese Ministry of the Environment (MOEJ a freezer in the laboratory of the Council for Scientific and Industrial (Ministry of Environment Japan), 2003), which targets the fraction of

1000

10000

100 1000

10 100 Cu by P-XRF (mg/kg) Co by P-XRF (mg/kg)

R2 = 0.966 R2 = 0.967 1 10 1 10 100 1000 10 100 1000 10000 HCl extractable Co by ICP-MS (mg/kg) HCl extractable Cu by ICP-MS (mg/kg)

100000 100000

10000 10000

1000 1000

100 100 Zn by P-XRF (mg/kg) Pb by P-XRF (mg/kg)

2 R = 0.719 R2 = 0.956 10 10 10 100 1000 10000 100000 10 100 1000 10000 100000 HCl extractable Zn by ICP-MS (mg/kg) HCl extractable Pb by ICP-MS (mg/kg)

Fig. 4. Comparison between the levels of metal(loid)s obtained through P-XRF and ICP-MS (after HCl extraction). The dotted line represents the 1:1 correlation. T. Itai et al. / Science of the Total Environment 470–471 (2014) 707–716 711

metal(loid)s leachable in gastric fluid. Three hundred milligrams of freeze-dried soil/ash mixture was put into a Teflon vessel with 10 mL of 1 M HCl and then shaken in darkness at 185 rpm in a reciprocating shaker for 2h. After filtration of extractant through a 0.45μm membrane filter (DISMIC, Advantec, Tokyo, Japan), concentrations of Mg, Al, V, Cr, Mn,Co,Cu,Zn,As,Sr,Cd,In,Sb,Cs,Ba,Pb,andBiweremeasuredusing an inductively coupled plasma mass spectrometer (ICP-MS; Agilent 7500cx, Agilent Technology, Tokyo, Japan). During measurement, instru- mental drift was calibrated using internal standards (89Yand103Rh). The precision of measurement was checked using triplicate measurement of the Montana soil SRM 2710a provided by the National Institute of Standards and Technology (USA); the precision was b2.5%. The color of soil/ash mixtures was measured using a portable soil color meter (SPAD-503, Konica-Minolta, Tokyo, Japan) after the samples were brought back to Japan. The ground samples were packed into an oxygen-impermeable film (AP-1826, ISO. Yokohama, Japan). A tungsten lamp was used as a light source for the soil color meter and reflected light from a sample placed in front of the window was detected by a silicon photocell. Object colors are expressed in a CIE1976 color space consisting of three axes, L*, a*, and b*. The L* value represents the lightness, which ranges from 0 (dark) to 100 (light). The a* and b* values represent the colors indicated along the red–green and yellow–blue axes, respectively. Both a* and b* values range from −60 to 60. The L*, a*, and b* values can be measured to a precision of one decimal place by this soil color meter. 367 60 N.A. 1.3 1.1 1.4 N.A. 35 1.1 73 N.A. 933 0.30 2.3. Statistical analysis 1100 All statistical analyses were performed with Statcel 2 (Seiun Inc., Tokyo, Japan). Spearman's rank correlation test was employed to assess the significance of correlations between various elements. Linear regression was used to evaluate the correlation between total metal(loid)s by P-XRF and HCl-extractable concentrations and correlations between various elements; the relationship between color coordinates (L*, a*, b*) and metal(loid) levels was studied using quadratic curves using Excel.

2.4. Noncancer risk assessment

The Hazard Quotient (HQ) was calculated for 13 metal(loid)s using the method recommended by the US Environmental Pro- tection Agency (US EPA, 2011). Only intake of the soil/ash mixture through ingestion was considered. The general equation for cal-

Mg Al V Cr Mn Fe Co Cu Zn As Br Sr Cd In Sn Sbculating Cs Ba HQ Hg is: Pb Bi and ICP-MS (HC1 extractable fraction). The unit of concentration is mg/kg dry weight. Color coordinates (L*, a*, b*) of each sample are also given. HCl 13 667 7.6 3.8 180 31,200 5.3 150 367 53 N.A. 2.4 1.2 N.D. N.A. 3.7 0.6 80 N.A. 157 N.D. HCl 14 146 2.3 4.4 44 17,000 1.5 123 93 40 N.A. 1.8 1.1 N.D. N.A. 3.8 0.6 26 N.A. 154 N.D. HCl 13 181 4.1 3.1 78 18,800 2.1 24 900 37 N.A. 2.2 1.0 N.D. N.A. 0.9 0.7 47 N.A. 50 N.D. HCl 18 400 2.7 1.7 36 19,400 467 27 36,700 39 N.A. 1.0 6.7 0.3 N.A. 2.5 2.8 20 N.A. 91 0.30 HCl 22 5000 4.0 8.6 84 43,000 19 5000 307 229 N.A. 4.2 8.4 8.0 N.A. 324 3.3 243 N.A. 6670 5.20 HCl 24 5670 3.7 8.8 91 28,000 11 4330 2770 128.0 N.A. 3.8 4.3 28 N.A. 325 4.5 280 N.A. 12,000 2.10 HCl 27 2700 5.7 5.1 116 36,600 50 1870 5000 99.0 N.A. 3.00 5.40 0.30 N.A. 76.0 1.20 140.0 N.A. 1390 0.90 HCl 19 17,700 5.3 14 119 41,400 24 6670 5000 253 N.A. 4.3 28 2.6 N.A. 392 4.1 210 N.A. 6000 2.10 HCl 12 1570 3.0 2.4 52 16,100 2.7 HCl 14 191 3.2 3.2 100 13,000 2.3C 28 IRS 120 34 EF N.A. 1.7ED 1.3 N.D. N.A. 2.1 0.5 43 N.A. 220 N.D. HQ ¼ oral 6 RfD0 BW AT 10 1.5 XRF N.A. N.A. N.A. N.A. N.D. 21,000 65 240 16,000 50 870 31 N.D. N.A. 55 N.D. N.A. N.D. N.D. 160 N.A. 1.3 XRF N.A. N.A. N.A. N.A. N.D. 17,000 4.0 12,000 13,000 730 730 150 N.D. N.A. 740 N.D. N.A. N.D. 150 11,000 N.A. 0.9 XRF N.A. N.A. N.A. N.A. N.D. 10,000 4.0 11,000 7100 N.D. 580 140 N.D. N.A. 790 N.D. N.A. N.D. 150 14,000 N.A. 0.9 XRF N.A. N.A. N.A. N.A. N.D. 27,000 6.0 22,000 19,000 1100 1500 140 N.D N.A. 510 N.D. N.A. N.D. 120 9800 N.A. − − − − where C is the concentration of HCl-extractable metal(loid)s in soil 0.1 2.2 XRF N.A. N.A. N.A. N.A. 150 35,000 9 4800 17,000 100 330 130 N.D. N.A. 220 N.D. N.A. N.D. 40 3100 N.A. 1.3 1.2 1.0 0.8 (μg/g), IRS is a conservative soil ingestion rate (adults: 100 mg/day) − − − − − (Calabrese et al., 1989; US EPA, 2011), EF is the exposure frequency 42.3 3.7 8.8 XRF N.A. N.A. N.A. N.A. 212 38,000 14 380 1200 N.D. 74 130 N.D. N.A. N.D. N.D. N.A. N.D. N.D. 360 N.A. 47.7 3.0 9.5 XRF N.A. N.A. N.A. N.A. N.D. 15,000 4 170 270 N.D. 72 100 N.D. N.A. N.D. N.D. N.A. N.D. N.D. 600 N.A. 45.7 5.1 9.7 XRF N.A. N.A. N.A. N.A. N.D. 26,000 8.0 55 220 N.D. 55 110 N.D. N.A. N.D. N.D. N.A. N.D. N.D. 100 N.A. 36.3 27.3 25.4 28.5 25.8 38.6 2.2 6.3 XRF N.A. N.A. N.A. N.A. N.D. 20,000 N.D. 3700 1300 N.D. 44 84 N.D. N.A. 1000 N.D. N.A. N.D. 25 1600 N.A. 44.3 6.0 13.0 XRF N.A. N.A. N.A.(365 N.A. N.D. 19,000 days/year), 5.0 53 ED 240 N.D.is the 1.8 110 exposure N.D. N.A. N.D. N.D. duration N.A. N.D. N.D. (adults: 380 N.A. 30 years), BW is the body weight (adults: 62 kg), AT is the average time of ″ ″ ″ ″ ″ ‴ ″ ″ ″ exposure (ED × 365 days/year), and RfD0 is the chronic oral reference ″ dose (mg/kg/day). RfD0s used in this study for metals are: V, 0.01; Cr, 33.63 37.39 34.87 36.70 37.02 36.10 36.02 35.43 32.09 ′ ′ ′ ′ ′ ′ ′ ′ ′ 33.15 ′ 0.001 (as hexavalent form); Mn, 0.01; Cu, 0.01; Zn, 0.03; As, 0.0003; Cd, 0°13 0°3 0°13 0°13 0°13 0°13 5°33 0°13 0°13 0°13 0.0005; Sb, 0.0005 (as total Sb) (Integrated Risk Information System by US EPA; www.epa.gov/iris); and Pb, 0.0036 (Healey et al., 2010). In the studied area, metal(loid) levels in soil/ash mixtures showed a very ″ 06.93 ″ 03.16 ″ 01.22 ″ 06.34 ″ 02.40 ″ 03.80 ″ 03.76 ″ 05.27 ″ 03.28 ″ 01.13 ′ large variation (described below). Thus, the exposure levels of these metal(loid)s may vary depending on the type of work done by the worker in the recycling area. People engaged in burning of e-waste or recovery of from residual ash might be exposed to higher levels of metals than those engaged in dismantling electric products, selling the product in the northeastern market, and so on. Hence, SampleGH10-E1 Latitude (N)′ 5°33 Longitude (W) L* a* b* GH10-E5GH10-E6′ 5°33 ′ 5°33 GH10-E2′ 5°33 GH10-E7′ 5°33 GH10-E8′ 5°33 GH10-E9′ 5°33 GH10-E10 5°33 GH10-E3′ 5°33 GH10-E4′ 5°33 Median (XRF)Median (HCl) N.A. N.A. 16 1119 N.A. N.A. 3.9 181 4.1 20,500 88 6.0 23,700 2040 8.2 4200 415 625 202 634 120 57 N.A. N.A. N.A. 625 2.3 N.A. 2.8 N.A. 2.0 N.A. N.A. N.A. 19 1100 N.A. 1.2 77 N.A. 577 1.5 Table 1 Concentration of 17 metal (loid)s in soil/ash mixtures determined by P-XRF N.D. Not detected. N.A. Not applicable (not measured). metal concentrations were determined for two different exposure 712 T. Itai et al. / Science of the Total Environment 470–471 (2014) 707–716

Common metals Rare metals 100000 1000

10000 100

1000

10

100

1 10 HCl extractable concentration (mg/kg dw.) HCl extractable concentration (mg/kg dw.) 1 0.1 Al Fe Cu Zn Cd Pb Cr Mn Co In Sr Ba As Sb Bi Transition Alkaline earth Metalloids metals metals

Fig. 5. Box-and-whisker plot of HCl-extractable metal(loid) concentration in soil/ash mixtures. The horizontal line indicates the median, the box covers the 25–75% percentiles, and the maximum length of each whisker is 1.5 times the interquartile range. Points outside this range show up as outliers.

scenarios and color was used as a proxy for high/low exposure. In ranges, as is the case for Fe. These results indicate that P-XRF is acceptable the high-exposure scenario, which covers people mainly engaged in to assess the distribution of metal(loid)s for a first screening, but open burning, the geometric mean concentrations of metal(loid)s in appropriate correction methods should be established before the results darksamples(E2,E3,E7,E8,andE9),whichshowedL*valuesb 37, are used in risk assessment. were used in calculating the oral intake via soil ingestion. The geometric mean concentrations of metal(loid)s in light samples 3.2. Element levels in soil (E1, E5, E6, and E10), which showed L* values N 42, were used in the moderate-exposure scenario, which covers people engaged in Of the 19 elements determined by HCl extraction, 10 showed tasks other than open burning. A Hazard Index (HI) was also variations of more than one order of magnitude (Al, Co, Cu, Zn, Cd, In, calculated as the sum of HQ values for all metals with an assumption Sb,Ba,Pb,andBi)(Fig. 5). Concentrations of Al, Cu, Zn, and Pb were that effects by all metal(loid)s were simply additive with no joint extremely high with maximum concentrations being of the order of toxic action. several thousand ppm. These metals were particularly high in dark samples such as E3, E8, and E9 (L*: 25.4–28.5), despite a slightly different 3. Results and discussion trend being observed for E7 (Table 1). Similar to common metals, highly toxic metal(loid)s (e.g., Cd and As) 3.1. Validation of P-XRF analysis and rare metal(loid)s (e.g., In, Sb, and Bi) also showed high concen- trations in dark samples (Table 1). Maximum levels of these rare The accuracy of P-XRF measurements was evaluated by measuring 18 metal(loid)s were comparable with the highest levels ever reported certified geological materials distributed by the Geological Society of from e-waste recycling sites (Table S3). Although the sources of these Japan. In most cases, the difference between the measured and certified metal(loid)s cannot be identified, they have various potential usages value was within one order of magnitude (Fig. 3). The determination in electrical products. Antimony trioxide is a major inorganic flame coefficient (R2) between the certified and measured values varied from retardant (Filella et al., 2002; Lu and Hamerton, 2002). Indium is used 0.602 to 0.997 (Fig. 3). Some elements (e.g., As, Sr) showed very good in flat panel displays and solar panels as a transparent conductive film agreement with the 1:1 line, whereas others showed systematic error. (In–Sn oxide), as well as in Pb-free solder (Ha et al., 2009; Robinson, For example, Mn, Cu, Zn, and Pb gave negative systematic errors in the 2009). Bismuth is widely used for low-melting alloys and metallurgical high concentration range. The approximate curve for Fe and Rb appears additives and is therefore common in electronic and thermoelectric to be parallel to the 1:1 line but has a positive y-intercept. Ti and appliances (Fowler et al., 2007). Ba clearly showed poor agreement between the certified and measured Although the P-XRF measurement data are only guide values, the values. high concentrations of Br (20–1500 mg/kg) and Hg (b20–150 mg/kg) Comparison between the P-XRF results and the HCl-extractable observed in dark samples are notable (Table 1). In this case, the source fractions determined by ICP-MS also showed good correlation (Fig. 4; of Br is possibly an organic form such as BFRs (Suzuki et al., 2009). Table 1). The determination coefficients (R2) between P-XRF and HCl- Incineration of organohalogen compounds with metals probably causes extractable levels were 0.966 for Co, 0.967 for Cu, 0.719 for Zn, and formation of dioxin-related compounds (DRCs) (Buser, 1986; McKay, 0.956 for Pb. As P-XRF measurements reflect bulk information, values 2002) and high concentrations of BFRs and DRCs were reported in this by P-XRF should be higher than the values of extractable metal(loid)s. site in a previous study (Bridgen et al., 2008). Hence, further monitoring The fraction of HCl-extractable phase relative to total level by P-XRF and risk assessment are necessary. varied from 0.27 to 5.56 for Co, 0.68 to 2.80 for Fe, 0.11 to 0.72 for Cu, 0.02 to 4.09 for Zn, and 0.26 to 0.86 for Pb. The HCl-extractable contents 3.3. Elemental correlations between Cu, Zn, Pb, and Al obtained for Co, Fe, and Zn that were greater than the P-XRF results should be attributed to interference by more abundant elements, or The elements with very high concentrations in the present study nonlinear response of fluorescence intensity in high concentration were Cu, Zn, Pb, and Al. By analyzing correlations between these metals, T. Itai et al. / Science of the Total Environment 470–471 (2014) 707–716 713

sources and behavior of these metals and minor metal(loid)s can be 0.593 0.699 0.377 evaluated. Cu, Pb, and Al were positively correlated (Table 2); Zn − indicates different behavior, mainly because of its higher concentration in E7. In an experimental study that examined the elemental composition of residual ash after burning circuit boards and insulated 0.136 0.461 − N.A. wire, Zn levels (180 and 764 mg/kg in circuit boards and insulated wire, respectively) were found to be lower than Cu (13,900 and 47,000 mg/kg) and Pb (3630 and 16,900 mg/kg) (Gullett et al., 2007). 0.001 N.A. 0.214 0.231 0.707 0.001 0.003 0.122 Hence, the present result indicates that other sources of Zn besides b b − circuit boards or insulated wire may also be contributing. Zn is used most commonly as a protective coating of other metals. Good correlation with Co, which is commonly added to Zn alloys to increase 0.721 0.666 0.593 0.304 0.019 0.035 0.071 0.558 melting point, mechanical strength, and resistance to oxidation, suggested that the predominant source materials for these were similar. Zn–Co alloys are commonly used as protective coatings 0.847 0.002 0.987 0.843 0.904 0.825 0.001 0.001 0.123 0.294 0.001 0.002 b b −

0.8. with a Co content generally between 0.6 and 1.0% (Pushpavanam, N 2000). Observed Co contents relative to Zn + Co ranged from 0.23 to 5.8% (geometric mean = 1.00%) and 0.03 to 3.51% (geometric 0.073 0.311 0.904 0.894 0.870 0.938 mean = 0.29%) by HCl extraction and P-XRF, respectively. Because − these values are close to the general Co content in alloys, Zn and Co in soil/ash mixtures are possibly supplied as protective coatings to other metals. 0.951 0.001 0.891 0.790 0.379 0.403 b Cu, Pb, and Al showed similar trends. However, the correlation between Cu and Pb was slightly lower, suggesting possible differences in sources of these two metals. As already described, the major 0.899 0.001 0.01 0.459 0.324 0.162 0.800 b

− products burned in this site are circuit boards and cables. There may be more Pb in circuit boards where it is mainly used as solder. If this is the case, the source of In, which showed the highest correlation 0.338 0.744 0.175 0.836 0.078 coefficient to Pb (R = 0.904, p = 0.013) rather than to Cu (R = 0.379, − p = 0.459) may also be because of its use as a solder. In, Sn and Bi are important metals in the production of Pb-free solder (Abtew and Selvaduray, 2000). The In level relative to Pb is still b0.3% in our 0.867 0.970 0.001 0.628 0.115 b

− study site but may increase in the future if the import of e-waste tractable concentration were used. Bold face indicates R value from developed nations continues. Al and Cu behave quite similarly so their sources may be similar. The good correlation 0.069 0.134 0.136 0.701 betweenAlandCd(R=0.951,pb 0.001) is curious but the reason − − − for this could not be identified.

3.4. Relationship between color of samples and metal(loid) contents 0.134 N.A. 0.001 0.850 0.001 0.339 0.014 b − To assess the risk of exposure, factors controlling spatial variation of

67 0.003 0.707metals 0.024 0.830should 0.005 be 0.248 examined. 0.013 Because used electrical products were 0.155 0.895 0.222 N.A. 0.001 0.711 N.A. 0.751 0.003 0.361 0.655 0.548 0.736 0.410 0.521 0.206 0.825 0.996

b burned in the central area of the recycling site, metal levels were highest − − − there. The color of a soil/ash mixture can be a useful indicator to assess the contribution of ash to the soil. The relationship between color coordinates (L*, a*, b*) and metal(loid) concentrations showed 0.087

− systematic correlations (Fig. 6). Among them, the L* value appears to be the best indicator of the residual ash:soil ratio because L* is a lightness factor. The determination coefficients (R2) of quadratic 0.399 curves between L* and Cu, Pb, As, and Sb were 0.962, 0.714, 0.934, − and 0.970, respectively; the value in E7 is an outlier because of its exceptionally high concentration of Zn. These correlations demonstrated that the sources of these metal(loid)s are clearly cance (p) between 19 elements and Cu. Zn, Pb, and Al. The data of HCI ex 0.001 0.475 0.034 0.670 0.001 0.598 0.006 0.538 N.A. 0.711 0.224 fi b b − the e-waste burned at this site, and the ground content is simply controlled by the dilution of residual ash by the surrounding soil. The poor correlation between Zn and L* suggests that the pre- 0.161 0.952 0.191 0.793 0.274 dominant source of Zn is different from that of Cu. − The good correlation between L* and metal(loid) levels suggests that the historical change in contamination can be followed if information on historical soil color distribution is available. The historical satellite data 0.069 0.001 0.656 0.895 b − provided by the Google Earth system showed some important changes cient (R) and statistical signi fi in the usage of the Agbogbloshie site (Fig. S1). In 2000, this area was

gl CrMnFeCoCuZnAsSeSrCdInSbCsBoPbBi MgAlV not used for massive e-waste recycling. The number of sheds in the northeastern part of the site was limited and the southern part of the area was covered by vegetation. The number of sheds gradually p 0.287 N.A. 0.498 Correlation with Al R 0.374 N.A. 0.243 0.931 0.256 0.669 Correlation with Cu R 0.591 p 0.054 0.071 0.998 0.01 1 0.789 0.101 0.5 p 0.072 p 0.612 0.850 0.444 0.534 0.253 0.812 Correlation with Pb R 0.625 0.593 0.001 0.759 0.097 0.549 Correlation with Zn R 0.183

Table 2 Correlation coef increased with time. A clear change was observed in the image of January 714 T. Itai et al. / Science of the Total Environment 470–471 (2014) 707–716

7000 20000

6000 Y = 10.72x2 -1051x 2 +25685 15000 Y = 25.49x -2331x 5000 R2 = 0.962 +53460 2 4000 R = 0.638 10000 3000

2000 5000 1000

0 0 HCI extractable CU in soil/ash mixture (mg/kg) HCI extractable 25 30 35 40 45 50 Al in soil/ash mixture (mg/kg) HCI extractable 25 30 35 40 45 50 L* L*

6000 14000

5000 12000 Y = 10.31x 2-1148x 10000 +30794 4000 R 2 = 0.714 8000 3000 6000 2000 4000

1000 2000

0 0

HCI extractable Zn in soil/ash mixture (mg/kg) HCI extractable 25 30 35 40 45 50 Pb in soil/ash mixture (mg/kg) HCI extractable 25 30 35 40 45 50 L* L*

300 400 Y = 0.47x 2-42.78x 350 Y = 0.89x 2-81.28x 250 +1013 +1879 300 2 2 200 R = 0.934 R = 0.970 250 150 200 150 100 100 50 50 0 0 HCI extractable As in soil/ash mixture (mg/kg) HCI extractable 25 30 35 40 45 50 Sb in soil/ash mixture (mg/kg) HCI extractable 25 30 35 40 45 50 L* L*

Fig. 6. Relationship between L* value and HCl-extractable concentration of Cu, Al, Zn, Pb, As, and Sb. The approximate quadratic curves and their equations are also given. The data of E7 were excluded from each plot because of its anomalously high content of Zn.

2008, when the open burning site could be clearly seen. The dark-colored those found at reference sites (Accra and Obuasi). The results of soil soil (probably soil/ash mixture) has remained since 2008. The change in analysis in the present study showed clearly elevated Pb and Sb levels the extent of dark soil during this period suggests a prominent change in in dark soil/ash mixtures, supporting high exposure of workers to cover soil. Such data suggest that at this location, at least since 2008, these toxic metal(loid)s. the workers are most probably exposed to high metal(loid) levels. The IncontrasttoPbandSb,theexposuresourceofFemayplausiblybe temporal change in the satellite image is consistent with previous different from that of inhalation/ingestion of cover soil, as the extractable reported data that point to the genesis of the Agbogbloshie scrapyard Felevelisnothighindarksoil/ashmixtures(Table 1). The typical daily in the early 1980s, and the emerging of e-waste scavenging activities Western European/American dietary intake of Fe was estimated to around 2006–2007 (Oteng-Ababio, 2012). be 15 mg/day. The estimated daily intake (DI) of Fe (ca. 2 mg/day) through soil/ash mixture is much less than this, even if we apply a 3.5. Comparison with urinary metal(loid) level conservative soil ingestion rate (100 mg/day) for the calculation of DI.Hence,significantly higher urinary Fe in e-waste recycling We previously measured the urinary metal(loid) levels of e-waste workers than in those at reference sites might have been caused by recycling workers in the Agbogbloshie site in 2008 (Table S4; Asante different factor(s). et al., 2012). Despite the different periods of sampling, comparison of Although Cu and Zn were extremely high in dark soil/ash soil levels with urinary levels would make sense because the recycling mixtures, their urinary levels in workers were not significantly higher activity has continued at this site. According to Asante et al. (2012), than those of normal residents in Accra and Obuasi (Asante et al., the urinary metal(loid) data after consideration of interaction by age 2012). Zn and Cu are essential elements for humans and their absorption suggested that Fe, Sb, and Pb levels were significantly higher than efficiencies through the gastrointestinal tract are homeostatically T. Itai et al. / Science of the Total Environment 470–471 (2014) 707–716 715

10 4. Conclusions HQ moderate HQ We have discussed the characteristics of metal(loid) contamination high 1 in an open burning site of e-waste in Ghana, as well as a health risk assessment. First, from a technical point of view, we assessed the applicability and limitation of P-XRF as a rapid screening tool for various metal(loid) levels in sites for open burning of e-waste. Because it is 0.1 difficult to transport large amount of soil from developing countries, on site method should be useful to evaluate geographical distribution of various metal(loid)s despite limited accuracy of this method. The 0.01 important outcomes of this monitoring are: (i) quite high levels of Al,

Hazard quotient Cu, Zn, and Pb in the open burning site; (ii) enrichment of rare metal(loid)s (e.g., Co, As, Sb, In, and Bi) was found in soil/ash mixtures; 0.001 (iii) there was good correlation of metal(loid) levels and soil color; and (iv) high HQ values of metal(loid) levels were found considering only the HCl-extractable fraction. The sources of some metal(loid)s could be characterized by analyzing correlations between elements. The 0.0001 main characteristic of this e-waste recycling site is the simultaneous V SeCrMnCoCdZnCuSbAsPb exposure to various contaminants including organic pollutants. Thus,

Fig. 7. Hazard quotient of 10 metal(loid)s. HQmoderate and HQhigh represent HQ values of the following approaches are required to assess human risk in e-waste moderate- and high-exposure scenarios, respectively. recycling sites more appropriately: (i) examination of the distribution of organic contaminants; (ii) speciation of metals in soil/ash mixture (particularly metal(loid)s for which tolerable intake values are given separately for their different chemical forms); (iii) a mechanistic controlled. Previous studies suggested that the absorption efficiency understanding of changes in chemical form(s) of organic/inorganic of Zn is concentration dependent, feedback controlled, and occurs contaminants during the open burning process; and (iv) risk assess- throughout the small intestine with the highest rate in the jejunum ment considering joint toxic action of metal(loid) mixtures as well as (Lee et al., 1989). Urine is possibly not a good indicator of Cu organic pollutants. Studies are now ongoing in our research laboratory exposure because ca. 98% of Cu excretion occurs through the bile on which we will report in the near future. and only the remaining 2% is excreted through urine (Wijmenga and Klomp, 2004). Hence, the exposure status of Zn and Cu cannot Conflict of interest be evaluated adequately using urine as a marker. Collecting data from other markers, such as blood and feces, is important. There is no conflict of interest regarding this work.

3.6. Risk assessment Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx. Workers in the e-waste recycling site of the present study are doi.org/10.1016/j.scitotenv.2013.10.037. exposed to soil/ash mixtures as they do not wear masks during the work. With the quite high levels of metals in soil/ash mixtures, inhalation/ingestion of the mixture must be an important References exposure route of toxic metals. Here, we present HQ estimation Abtew M, Selvaduray G. Lead-free solders in microelectronics. Mater Sci Eng R: Rep values when only ingestion of soil and a conservative soil ingestion 2000;27:95–141. rate were used for the calculation of daily ingestion rates and HQ Amoyaw-Osei Y, Agyekum O, Pwamang J, Mueller E, Fasko R, Schluep M. Ghana e-waste (Fig. 7). country assessment. SBC e-Waste Africa Project; 2011111. Asante KA, Agusa T, Biney CA, Agyekum WA, Bello M, Otsuka M, et al. Multi-trace element For the moderate-exposure scenario, the HQ value for any single levels and arsenic speciation in urine of e-waste recycling workers from metal(loid) exposure does not exceed unity. However, HQs of Pb Agbogbloshie, Accra in Ghana. Sci Total Environ 2012;424:63–73. and As exceeded 0.1 and the HI value in this scenario is N1.0. Hence, Atiemo SM, Ofosu FG, Aboh IK, Kuranchie-Mensah H. Assessing the heavy metals contamination of surface dust from waste electrical and electronic equipment even in the moderate-exposure scenario, the workers are at risk of (e-waste) recycling site in Accra, Ghana. Res J Environ Earth Sci 2012;4:605–11. metal(loid) exposure. The hazard level in the high-exposure scenario Bridgen K, Labunska I, Santillo D, Johnston P. Chemical contamination at e-waste recycling is surely much worse than in the moderate scenario. HQs for and disposal sites in Accra and Korforidua, Ghana. Amsterdam: ; 2008. Buser HR. Polybrominated dibenzofurans and dibenzo-p-dioxins: thermal reaction exposure to Cu, Sb, As, and Pb exceeded unity with Pb having products of polybrominated diphenyl ether flame retardants. Environ Sci Technol the highest HQ of 5.4. Zn and Cd indicated HQ N 0.1 and the HI 1986;20:404–8. valueinthisscenariois9.5.Becausetheexposurestatustothese Calabrese EJ, Barnes R, Stanek EJ, Pastides H, Gilbert CE, Veneman P, et al. How much soil do young children ingest: an epidemiologic study. Regul Toxicol Pharm 1989;10: metal(loid)s is at serious levels, remediation of the cover soil and 123–37. improvement of the working environment in the present study Caravanos J, Clark E, Fuller R, Lambertson C. Assessing worker and environmental site are required. chemical exposure risks at an e-waste recycling and disposal site in Accra, Ghana. – Some metals, e.g., Sb, As, and Cr have different RfD values depending Blacksmith Inst J Health Poll 2011;1:16 25. 0 Filella M, Belzile N, Chen Y-W. Antimony in the environment: a review focused on natural on their chemical form. Because information on the chemical forms of waters: I Occurrence. EarthSci Rev 2002;57:125–76. these metal(loid)s at the e-waste recycling site is limited, speciation Fowler BA, Sexton MJ, Bismuth. Handbook on the toxicology of metals. In: Nordberg GF, – study of the solid phase will be an important focus in e-waste recycling Fowler BA, Nordberg M, Friberg LT, editors. 3rd ed. Academic Press; 2007. p. 433 43. Gullett BK, Linak WP, Touati A, Wasson SJ, Gatica S, King CJ. Characterization of air sites. In addition, RfD0 values of some rare metal(loid)s such as In and Bi emissions and residual ash from open burning of electronic wastes during simulated have not been established. As enrichment of these metals in the e-waste rudimentary recycling operations. J Mater Cycles Waste Manag 2007;9:69–79. recycling site has recently been reported as a matter of concern (Ha Ha NN, Agusa T, Ramu K, Tu NP, Murata S, Bulbule KA, et al. Contamination by trace elements at e-waste recycling sites in Bangalore, India. Chemosphere 2009;76:9–15. et al., 2009; Sjögren et al., 2007), toxicological studies of these emerging Healey N, Jones-Otazo H, Walker M, KnaflaA.Toxicological review and recommended contaminants should be addressed. toxicological reference values for environmental lead exposure in Canada. Report 716 T. Itai et al. / Science of the Total Environment 470–471 (2014) 707–716

prepared for Health Canada, Healthy Environments and Consumer Safety Branch. Sjögren B, Iregren A, Elinder C-G, Yokel RA. Aluminum. In: Nordberg GF, Fowler BA, Ottawa, ON: Environmental Health Bureau; 2010. Nordberg M, Friberg LT, editors. Handbook on toxicology of metals. Academic HuoX,PengL,XuX,ZhengL,QiuB,QiZ,etal.Elevated blood lead levels of children in Guiyu, Press; 2007. p. 339–52. an electronic waste recycling town in China. Environ Health Perspect 2007;115:1113–7. Suzuki G, Kida A, Sakai S-i, Takigami H. Existence state of bromine as an indicator of the MOEJ (Ministry of Environment Japan). Notification Concerning Soil Pollution; 2003 [No. source of brominated flame retardants in indoor dust. Environ Sci Technol 2009;43: 19, in Japanese]. 1437–42. Lee HH, Prasad AS, Brewer GJ, Owyang C. Zinc absorption in human small intestine. Am J Tanabe S. Environmental specimen bank in Ehime University (es-BANK), Japan for global Physiol 1989;256:G87–91. monitoring. J Environ Monit 2006;8:782–90. Leung A, Cai ZW, Wong MH. Environmental contamination from electronic waste Terazono A, Murakami S, Abe N, Inanc B, Moriguchi Y, Sakai S-i, et al. Current status recycling at Guiyu, southeast China. J Mater Cycles Waste Manag 2006;8:21–33. and research on E-waste issues in Asia. J Mater Cycles Waste Manag 2006;8: Leung AO, Zheng J, Yu CK, Liu WK, Wong CK, Cai Z, et al. Polybrominated diphenyl ethers 1–12. and polychlorinated dibenzo-p-dioxins and dibenzofurans in surface dust at an Tue NM, Sudaryanto A, Minh TB, Isobe T, Takahashi S, Viet PH, et al. Accumulation of e-waste processing site in Southeast China. Environ Sci Technol 2011;45:5775–82. polychlorinated biphenyls and brominated flame retardants in breast milk from Li J, Duan H, Shi P. Heavy metal contamination of surface soil in electronic waste women living in Vietnamese e-waste recycling sites. Sci Total Environ 2010;408: dismantling area: site investigation and source-apportionment analysis. Waste 2155–62. Manag Res 2011;29:727–38. US-EPA. Field portable X-ray fluorescence spectrometry for the determination of Lu S-Y, Hamerton I. Recent developments in the chemistry of halogen-free flame elemental concentrations in soil and sediment. Available at http://www.epa.gov/, retardant polymers. Prog Polym Sci 2002;27:1661–712. 2007. [Verified 18 Mar 2013]. McKay G. Dioxin characterisation, formation and minimisation during municipal solid US-EPA. Exposure factors handbook: 2001 edition. Available from theSpringfield, VA. waste (MSW) incineration: review. Chem Eng J 2002;86:343–68. Washington, DC: National Technical Information Service; 2011 [EPA/600/R-09/052F]. Oteng-Ababio M. When necessity begets ingenuity: scavenging for survival in a US-EPA. Integrated Risk Information System (IRIS). Available at www.epa.gov/iris. globalizing city. Afr Stud Q 2012;13:2152–448. [Verified 18 Mar 2013]. Pushpavanam M. Critical review on alloy plating: a viable alternative to conventional Widmer R, Oswald-Krapf H, Sinha-Khetriwal D, Schnellmann M, Böni H. Global plating. Bull Electrochem 2000;16:559–66. perspectives on e-waste. Environ Impact Assess Rev 2005;25:436–58. Robinson BH. E-waste: an assessment of global production and environmental impacts. Wijmenga C, Klomp LW. Molecular regulation of copper excretion in the liver. Proc Nutr Sci Total Environ 2009;408:183–91. Soc 2004;63:31–9.