at SciVerse ScienceDirect

Environmental Pollution 180 (2013) 259e264

Contents lists available

Environmental Pollution

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

Historical arsenic contamination of soil due to long-term phosphatefertiliser applications

Tom N. Hartley a, Andy J. Macdonald a, Steve P. McGrath a, Fang-Jie Zhao b,*

aRothamsted Research, Harpenden, Hertfordshire AL5 2JQ, UKbCollege of Resources and Environmental Sciences, Nanjing Agricultural University, Nanjing 210095, China

a r t i c l e i n f o

Article history:Received 2 April 2013Received in revised form18 May 2013Accepted 21 May 2013

Keywords:ArsenicHerbagePhosphate fertiliserSoil contamination

* Corresponding author.E-mail addresses: Fangjie.Zhao@njau.edu.cn, Fa

(F.-J. Zhao).

0269-7491/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.envpol.2013.05.034

a b s t r a c t

Archived samples from the Park Grass Experiment, established in 1856, were analysed to determine theimpacts of long-term phosphate fertiliser applications on arsenic concentrations in soil and herbage. Inplots receiving 35 kg P ha�1 annually (þP), topsoil As concentrations almost doubled from an initial valueof w10 mg kg�1 during 1888e1947 and remained stable thereafter. The phosphate fertilisers used before1948 contained 401e1575 mg As kg�1, compared to 1.6e20.3 mg As kg�1 in the later samples. Herbagesamples from the þP plots collected during 1888e1947 contained significantly more As than those fromthe �P plots, but later samples did not differ significantly. Mass-balance calculations show that the in-crease in soil As can be explained by the As input from P fertiliser applications before 1948. The resultsdemonstrate that the P fertilisers used on the Park Grass Experiment before 1948 caused substantial Ascontamination of the soil.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Phosphate fertilisers provide one of the most important nutri-ents for crop production. Globally about 150 million tonnes ofphosphate rock are extracted each year for the production ofphosphate fertilisers and the demand for this finite resource isprojected to increase to feed a growing global population (Cordellet al., 2009; Van Vuuren et al., 2010). However, P fertilisers can bea source of toxic trace elements to agricultural soils, the best knownexample being cadmium (Chen et al., 2007; Grant and Sheppard,2008; Jiao et al., 2012; McLaughlin et al., 1999; Nziguheba andSmolders, 2008). Trace element concentrations in phosphate rockvary greatly depending on the type and source of the mineral(Chaney, 2012; Van Kauwenbergh, 1997). Sedimentary rocks oftencontain more impurities than their igneous counterparts (Staceyet al., 2010) and also make up the bulk of global P fertiliser pro-duction (Cordell et al., 2009; Stacey et al., 2010). As clean phosphatereserves are depleted it is likely that phosphate rocks with higherconcentrations of toxic trace elements will increasingly be used.This will almost certainly increase the risk of soil contamination(Jiao et al., 2012; Stacey et al., 2010) because there are currently no

ngjie.Zhao@rothamsted.ac.uk

All rights reserved.

economic methods for removing contaminants in phosphate rocksduring the manufacture of P fertilisers (Stacey et al., 2010).

Arsenic (As) is a widely occurring environmental contaminantwith inorganic As recognised as a class-one carcinogen (NationalResearch Council, 2001). Arsenic in the soil is derived from boththe parent materials and anthropogenic sources such as mining,smelting, the use of As-containing pesticides and animal manures,and irrigation of As-contaminated groundwater (Meharg and Zhao,2012). Arsenic contamination in soil may cause phytotoxicity andyield losses (Panaullah et al., 2009). Its entry into the food chain viaplant uptake can also pose a significant risk to human health,especially for rice consumers, as paddy rice is prone to As accu-mulation (Meharg et al., 2009; Meharg and Zhao, 2012). It istherefore necessary to understand the sources of the contaminant,its transformation in soil and uptake by plants.

Inorganic fertilisers are not generally considered to be animportant source of As contamination. Several recent studies havereported relatively low concentrations of As in P fertilisers.Nziguheba and Smolders (2008) analysed 196 P fertilisers used in12 European countries and obtained a mean As concentration of7.6 mg kg�1 and a 95 percentile value of 21 mg kg�1. Jiao et al.(2012) summarised the data for the USA, showing that 93% of thesamples (n ¼ 203) contain <20 mg As kg�1 with only 2 samplesexceeding 60 mg As kg�1. The ranges reported for China and Chileare also comparable at 0e58 and 8e20 mg As kg�1, respectively(Jiao et al., 2012). Therefore, based on the average application rate

Table 1Fertiliser and lime treatments on selected plots of the Park Grass Experiment andcorresponding soil (0e23 cm) pH values, as measured in 2011.

Plot/sub-plot Fertiliser# and liming regime Soil pH in H2O

3a None, limed 7.23d None, unlimed 5.37a P, K, Na, Mg, limed 7.07d P, K, Na, Mg, unlimed 4.99/2a N2, P, K, Na, Mg, limed 7.19/2d N2, P, K, Na, Mg, unlimed 3.714/2a N*2, P, K, Na, Mg, limed 7.014/2d N*2, P, K, Na, Mg, unlimed 6.018a (N2), (P), (K), (Na), (Mg), limed 7.118d (N2), (P), (K), (Na), (Mg), unlimed 3.9

Note: plots 3, 7 and 9 started in 1856, plot 14 in 1858 and plot 18 in 1865.#Rates of fertiliser, applied annually.P, 35 kg P ha�1 as superphosphate made from bone ash (1856/8e1888), single su-perphosphate (1889e1896, and 1903e1986), basic slag (1897e1902) or triple su-perphosphate (since 1987).(P), 4.4 kg P ha�1 as superphosphate made from bone ash (1865e1888), single su-perphosphate (1889e1896, 1903e1904) or basic slag (1897e1902).K, 135 kg K ha�1 as potassium sulphate (1856/8e1878) or 225 kg K ha�1 (since1879).(K) 42 kg K ha�1 as potassium chloride (1865e97), 34 kg K ha�1 as potassium sul-phate (1898e1904) or 225 kg K ha�1 (since 1905).Na, 31 kg Na ha�1 as sodium sulphate (1856/58e1863) or 15 kg Na ha�1 (since1864).(Na), 21 kg Na ha�1 (1865e1870) or 42 kg Na ha�1 (1871e1904) as sodium silicate,or 15 kg Na ha�1 as sodium sulphate (since 1905).Mg, 10 kg Mg ha�1 as magnesium sulphate since 1856/58.(Mg) 4 kg Mg ha�1 (1865e1904) or 10 kg Mg ha�1 (since 1905).N2, 96 kg N ha�1 as ammonium sulphate since 1856; (N2) 39 kg N ha�1 (1865e1904) as ammonium sulphate or 96 kg N ha�1 (since 1905).N*2, 96 kg N ha�1 as sodium nitrate since 1858.Regular liming began in 1903 on parts of plots 3, 7 and 9; and in 1920 on plots 14and 18.

T.N. Hartley et al. / Environmental Pollution 180 (2013) 259e264260

of P fertilisers and the mean As concentration, the inputs of As via Pfertilisers are small and unlikely to lead to a significant accumula-tion in the soil (Chen et al., 2007; Jiao et al., 2012; Nziguheba andSmolders, 2008). However, some igneous phosphate rocks inSweden are reported to contain high As concentrations (VanKauwenbergh, 1997).

The long-term experiments at Rothamsted, and the associatedSample Archive, have been used to identify changes in the sourcesand amounts of heavymetals, including Cd and Pb, in soil (e.g. Joneset al., 1992; Jones and Johnston, 1991; Nicholson et al., 1994). In thepresent study, we used the Park Grass Experiment at Rothamsted,which has been running continuously for over 150 years, toexamine the long-term effect of P fertiliser applications on Ascontamination of the soil.

2. Materials and methods

2.1. The Park Grass Experiment

The Park Grass Experiment was established in 1856 to test the effect of mineralfertilisers and organic manures on hay production (Silvertown et al., 2006). Theexperiment is located on a site of approximately 2.8 ha on the Rothamsted estate,Hertfordshire, UK. The site had been under grassland for at least a century prior tothe start of the experiment; the vegetation was initially uniform. The soil is aChromic Luvisol (FAO classification); the topsoil (0e23 cm) is a slightly flinty silty-clay loam (Anon, 2006; Blake et al., 1999). Treatments include unfertilised con-trols and plots with different combinations and amounts and of N, P, K, Mg and Na.Nitrogen is applied either as ammonium sulphate or sodium nitrate. The plots arecut in mid-June to make hay. In the early years of the experiment the regrowth wasgrazed, but this stopped in 1875 (Silvertown et al., 2006); since then a second cut hasbeen taken and removed green in autumn (SeptembereNovember). Before 1960,hay yields were determined on plant samples collected from the field after cuttingand turning; hay samples were chopped and dried before archiving. From 1960onward herbage yields in June have been determined on fresh material taken fromstrips using a forage harvester. The freshly cut samples are chopped and dried at80 �C prior to archiving. The remainder of the plot is cut andmade into hay as before(Kohler et al., 2010). Most plots were split into limed and non-limed halves in theearly twentieth century and were further sub-divided into four sub-plots (a, b, c andd) in 1965. The “a”, “b” and “c” sub-plots are limed to maintain soil (0e23 cm) pHclose to 7, 6 and 5, respectively; the “d” sub-plot is unlimed. Both limed and unlimedsub-plots, with and without phosphate fertiliser applications, were selected for thisstudy; sub-plots included 3a, 3d, 7a, 7d, 9/2a, 9/2d, 14/2a, 14/2d, 18a, 18d (Table 1).

2.2. Archived samples of soil, herbage and fertilisers

In the earlier years of the experiment, soil was sampled to a depth of 23 cm usinga 15 � 15 cm or a 30 cm � 30 cm open-ended metal box, with samples taken fromone, two or three locations within each plot. Since 1923, gouge augers of 2e3 cmdiameter have been used to collect about 18 soil cores at random within each sub-plot. Soil cores are bulked and broken up by hand to remove obvious plant material,stones and soil fauna (e.g. worms) prior to air-drying, chemical analysis and storagein the Rothamsted Sample Archive. Soil samples (0e23 cm), spanning the periodfrom 1870 to 2011, were obtained from the archive for each sub-plot; between 13and 16 samples, each of about 3 g, were collected in each case. In addition, toinvestigate the distribution of P and As in the soil profile, samples collected in 2011from the 0e23, 23e46 and 46e69 cm depths of plots 3a, 3d, 9a and 9d were alsotaken for analysis.

Sixteen or seventeen herbage samples (Cut 1 only), each of about 0.5 g, wereobtained from the Sample Archive for each of the selected sub-plots. Wheneverpossible the herbage samples corresponded to the years when soil was also sampled.Herbage yield datawere retrieved from the Electronic Rothamsted Archive (e-RA). Inaddition, 29 samples of phosphate fertiliser, each of about 1 g, used on Park Grassand other long-term experiments from 1925 to 2007 were retrieved for analysis.

2.3. Analytical methods

Soils were milled to <40 mesh in a Retsch PM400 Milling Machine at 250 rpmfor 6 min, whilst fertilisers were ground using a clean pestle and mortar. Sampleswere dried in an oven at 80 �C overnight prior to aqua regia digestions. The methodof digestion was based on McGrath and Cunliffe (1985). Soil or fertiliser sample(0.25 g) was digested with 5 ml of aqua regia (4 ml concentrated hydrochloric acidand 1 ml concentrated nitric acid, both of the high purity grade) in a programmedheating block. Every fifth sample was repeated for quality control. In each digestionbatch, two blanks and a certified referencematerial (NIST 2711eMontana soil) wereincluded for quality assurance. The digest solutions were analysed by InductivelyCoupled Plasma Optical Emission Spectrometry (ICP-OES, Perkin Elmer Optima

7500DV) for P and other major elements and Inductively Coupled Plasma MassSpectrometry (ICP-MS, Perkin Elmer NexION 300X) for As. NIST 2711 has a certifiedtotal As concentration of 105� 8 mg kg�1, and a concentration of 93.1 �3.7 mg kg�1

(89% recovery) was obtained. Average CV% for repeated soil analysis was 1.3%.Herbage samples were dried in an oven at 80 �C overnight prior to digestion.

Between 0.25 and 0.50 g of plant material was digested with 5 ml of nitric acid/perchloric acid (87:13 v/v, both of the high purity grade) (Zhao et al., 1994); everyfifth sample was repeated. Each batch of digestion included two blanks and acertified reference material (NIST 1570a e spinach leaves) for quality assurance.The digest solutions were analysed by ICP-OES and ICP-MS for P and As,respectively. Repeated analysis of NIST 1570a gave a mean As concentration of0.058 � 0.023 mg kg�1, compared with the certified concentration of0.068 � 0.012 mg kg�1 (i.e. 85% recovery). Average CV% for repeated herbageanalysis was 13.5%.

2.4. Solidesolution partition of As

The partitioning of As between the solid and solution phases of the soil wasdetermined to provide an estimate of potential As leaching. In November 2011,approximately 30 cores of soil from the 0e23 cm layer were taken from the un-treated areas around the edge of the Park Grass (so as not to disturb the experiment)using a 3 cm diameter gouge auger. These soil cores were bulked, air-dried andsieved through a 5 mm mesh. The areas surrounding the experiment have notreceived significant fertilisers or lime inputs. Therefore, the soil from these areas isconsidered to be similar to that from plot 3d (Table 1). Twenty grams of soil wasplaced in a 50 ml centrifuge tube and acidified with 0, 2, 4, 6, 8, 10, 12, 14, 16, 18 or20 ml of 0.1 M HNO3, with two replicates for each level of acidification. Each tubewas made up to 20 ml with deionized water (>18.2 MU). The soil/liquid suspensionswere shaken at 150 rpm in an incubator shaker at 21 �C. A similar incubationexperiment was set up for 6-weeks with occasional shaking. Samples were thencentrifuged at 3000 g for 30 min and the supernatants filtered through 0.45 mmsyringe filters. The solutions were acidified to 2% HNO3 prior to ICP-MS analysisfor As.

2.5. Data analysis

The Park Grass Experiment was set up well before the advent of modern sta-tistics; consequently the experimental design does not include replications of in-dividual fertiliser treatments. Therefore, in order to assess the effect of P fertilisers

0

200

400

600

800

1000

1200

1400

1600

1800

1870 1890 1910 1930 1950 1970 1990 2010

P C

once

ntra

tion

(mg

kg-1

)

Year

3a3d7a7d9/2a9/2d14/2a14/2d18a18d

0

5

10

15

20

25

1870 1890 1910 1930 1950 1970 1990 2010

As C

once

ntra

tion

(mg

kg-1

)

Year

3a3d7a7d9/2a9/2a14/2a14/2d18a18d

Fig. 1. Changes in the concentrations of P (A) and As (B) in the topsoil (0e23 cm) from

T.N. Hartley et al. / Environmental Pollution 180 (2013) 259e264 261

on herbage As concentration in the different �P or þP treatments data collected inmultiple years during the study period (1870e2011) were used as replicates inanalysis of variance (ANOVA) using Genstat version 14 (VSN International, HemelHempstead, UK).

To investigate whether the As contaminated P fertilisers could give rise to theobserved increase in soil As concentration, a Monte Carlo type simulation waswritten using the Python 2.7.3 programming language (http://www.python.org/,accessed on 25th April 2012). A mass-balance approach was employed whichaccounted for the As inputs from P fertilisers and atmospheric deposition, and theoutputs by potential leaching and plant off-take. The initial soil As concentrationwastaken to be between 9.1 and 10.5 mg kg�1 based on experimental data. The bone ashused to make superphosphate was the initial P source prior to single superphos-phate (SSP), made frommineral phosphate, which was first applied in 1889 (Warrenand Johnston, 1964), and the assumption that As contaminated P fertilisers wereapplied from 1889 is reflected in the simulation. Bone ash is assumed to have anegligible As concentration based on the findings of Garcia and Rosentrater (2008).The As input from P fertilisers was obtained by pseudo-randomly generating anumber from a range based on observed As concentrations in the fertiliser samplesused between 1925 and 1947 and between 1948e2007. The concentration wasmultiplied by the mass of P fertiliser applied per year (439 kg ha�1 for SSP and171 kg ha�1 for triple superphosphate which replaced SSP from 1987). Basic slag wasused instead of SSP during 1897e1902, and is assumed to contain no As. Total at-mospheric deposition of As is taken to be 2e3 g ha�1 yr�1 based on the data pub-lished by the Department for Environment and Rural Affairs of the UK (http://pollutantdeposition.defra.gov.uk/node/397, accessed on 15th May 2012). Removalof As by plant uptake was 0.025e3 g ha�1 yr�1 (excluding 2 outlier values) based onthe measured As concentrations in herbage and known hay yields, with theassumption that the herbage from Cut 2 has As concentrations in the same range asin Cut 1. Deviation from this assumption would not significantly affect the outcomeof simulations because the yield of Cut 2 is generally small (on average approxi-mately one third of the yield of Cut 1) (Anon, 2006) and herbage As off-take is alsosmall compared with other As inputs and outputs. Leaching of As to lower horizonswas calculated from the typical annual drainage of 250mm for the experimental site(Knights et al., 2000) and the As concentration in solution measured in the 24 hsolidesolution partition experiment. The bulk density of soil was modelled as being0.95e1.30 g cm�3 based on Hopkins et al. (2009) and unpublished data. All input andoutput data were pseudo-randomly generated from uniform distributions boundedby the ranges described above. The simulation was run 500,000 times.

different plots of the Park Grass Experiment. See Table 1 for treatment information.

3. Results

3.1. Trends of P and As concentrations in soil

In plot 3, to which P fertilisers have never been applied, soil Pconcentration (as determined by the aqua regia digestion) declinedslightly (significant at P < 0.001) over the w150 year period(Fig. 1A). Plot 18 received a small amount of P (4.6 kg P ha�1 yr�1) inthe early years between 1865 and 1904 (Table 1), but this did notincrease soil P concentration and, for this reason, plot 18 isconsidered to be �P in the subsequent discussion. In plots 7, 9 and14 to which 35 kg P ha�1 has been applied annually, soil P con-centration increased steadily, reaching a current level that is 2e3times higher than that in plots 3 and 18. In the unlimed section ofplot 9 (9/2d), the increase in soil P concentration was greater thanthat in plot 14 (Fig. 1A), possibly due to a smaller plant uptake in theformer due to soil acidification caused by the long-term use ofammonium sulphate. When lime was added, the three þP plotsshowed similar trends of increasing P concentration (Fig. 1A).

The concentration of As in the topsoil was around 10 mg kg�1 in1870, soon after the experiment had begun, and has remainedstable in the �P plots (Fig. 1B). In contrast, all three þP plotsshowed substantial increases in soil As concentration during theperiod from the mid-1880s to the late 1940s, and since then theconcentration has remained stable at approximately double thelevel found in the�P plots. In both the limed and unlimed sections,the trends of soil As concentration were similar.

3.2. Arsenic concentration in the archived P fertilisers

The P fertilisers used from 1925 to 1947 had very high con-centrations of As, ranging from 401 to 1575 mg kg�1 (Fig. 2A). From

1948 onwards, As concentration in P fertilisers dropped to between0 (undetectable) and 13mg kg�1. This pattern is consistent with thetrend of increasing As concentration in the soils from the þP plotsbefore 1948 but not since. Single superphosphate had been used inthe experiment before 1986, since then it has been replaced withtriple superphosphate. This is reflected in the increase in the Pconcentration of the fertilisers (Fig. 2B) from 1986. However, the Asconcentrations in the P fertilisers decreased sharply well before theswitch from single to triple superphosphate.

Based on the As concentrations determined in the archived Pfertiliser samples, the amounts of As added in the P fertilisersvaried from 175 to 689 g ha�1 yr�1 (mean 373 g ha�1 yr�1) during1925e1947 and from 0.7 to 6.2 g ha�1 yr�1 (mean 4.0 g ha�1 yr�1)since 1948. To calculate if the additions of P fertilisers could accountfor the increases in soil As in the þP plots, we used Monte Carlosimulations with the range of As concentrations (401e1575 mg kg�1) found in the P fertilisers used between 1925 and1947. Based on the pattern of soil As (Fig. 1B), we assumed that Pfertilisers containing high As concentrations were used during theperiod between 1888 and 1947 (except 1897e1902, when basic slagwas used); before 1888 bone ash was used to make P fertilisers,which were highly unlikely to contain much As (A. E. Johnston,personal communication). In the simulations, As inputs included Pfertilisers and atmospheric deposition, whilst As outputs includedplant uptake and leaching; both the inputs and outputs wereassumed to affect As concentration in the topsoil (0e23 cm) only.The results of simulation are in excellent agreement with theobserved increases in the soil As concentration, with most of theobserved data points lying around the average and within theminimum andmaximum of the simulations (Fig. 3), suggesting thatthe assumptions used in the simulation were valid. Simulations

Fig. 4. Distribution of As (A) and P (B) concentrations in the soil profile of plots 3 and9/2 sampled in 2011.

Fig. 2. Concentrations of As (A) and P (B) in the archived phosphate fertilisers used inthe Park Grass Experiment. SSP, single superphosphate; TSP, triple superphosphate.

T.N. Hartley et al. / Environmental Pollution 180 (2013) 259e264262

were also run for plots 3 and 18, and the results show either nochange in soil As concentration in plot 3 or a small increase ofapproximately 0.2 mg As kg�1 in plot 18 due to the applications ofsmall amounts of P in the form of high-As SSP in the early years(data not shown).

3.3. Distribution of As in the soil profile

Soil samples taken from different depths of plots 3 and 9/2 in2011 were analysed for the distribution of As and P (Fig. 4). In both

0

5

10

15

20

25

30

1860 1880 1900 1920 1940 1960 1980 2000

As C

once

ntra

tion

(mg

kg-1

)

Year

Average

Min

Max

7a

7d

9/2a

9/2d

14/2a

14/2d

Fig. 3. Monte Carlo simulations of the soil As concentrations in the þP plots of the ParkGrass Experiment. Lines represent the range and the average of simulation results;symbols represent observed values in different þP plots.

the unlimed and limed sections of plot 3, there was little variationin the As concentration from 0 to 69 cm with an average of9 mg kg�1. In plot 9/2d (þP), the topsoil (0e23 cm) contained aboutdouble the As concentration of that in plot 3, whereas the subsoil(23e46 and 46e69 cm depths) had similar concentrations to thoseobserved in plot 3. In the limed section of plot 9/2 (a), As concen-tration in the 46e69 cm depth was larger than that in the unlimedsection of plot 9/2 (d) or in plot 3, suggesting some possiblemovement of As down the soil profile. Liming also increased the Pconcentration in the subsoil from the 46e69 cm depth while itslightly decreased that in the topsoil in plot 9/2.

3.4. Solubility of As in soil and potential leaching losses

Topsoil collected from the edge of the experiment was extractedwith water, with or without additions of HNO3. After equilibrationfor 1 day, the As concentration in the solution phase from the un-acidified suspension (pH 5.0) was approximately 20 mg L�1; thisdecreased to between 7 and 10 mg L�1 in the acidified suspensions(pH 2.5e4.4) (Fig. 5). The As partition coefficient (Kd ¼ total soil Asconcentration inmg kg�1/soil solution As inmg L�) varied from 440to 1300 L kg�1, similar to the range reported for Californian crop-land soils (Chen et al., 2009). Assuming an annual drainage of250 mm which is typical of the experimental site (Knights et al.,2000), leaching of As could amount to 18e50 g ha�1, accounting

pH2 3 4 5 6 7

As

conc

entra

tion

(mg

L-1)

0.00

0.02

0.04

0.06

0.08

0.10

0.12

1 day6 weeks

------Soil acidified-----

--Soil not acidified--

Fig. 5. The concentration of As in the soil solution after incubation for 1 day or 6weeks.

0.000.050.100.150.200.250.300.350.400.45

-P +P -P +P

1888-1947 1959-2011

As c

once

ntra

tion

(mg

kg-1

) A

0

500

1000

1500

2000

2500

3000

3500

-P +P -P +P

1888-1947 1959-2011

P co

ncen

tratio

n (m

g kg

-1) B

Fig. 6. The As (A) and P (B) concentrations in herbage from the �P (plots 3 and 18)and þP (plots 9, 9/2 and 14/2) plots during the 1888e1947 and 1959e2011 periods.*Denotes significant (P < 0.05 difference between the �P and þP plots.

T.N. Hartley et al. / Environmental Pollution 180 (2013) 259e264 263

for 0.07e0.2% of the total As in the topsoil. When the soilewatersuspension was incubated for 6 weeks, pH increased comparedwith the values measured after only 1 day of equilibration (Fig. 5).In the un-acidified suspension, the increase in pH (to 6.3) was mostlikely to be due to reduction reactions occurring in the soil, whichconsumed protons. In this case the concentration of As in soil so-lution increased markedly to approximately 100 mg L�1 (Fig. 5),possibly due to arsenate reduction and the subsequent increase inarsenite solubility (Masscheleyn et al., 1991; Xu et al., 2008;Yamaguchi et al., 2011); in this case annual As leaching lossescould amount to w1% of the total As in the topsoil (0e23 cm).When the soil was acidified, even a 6-week incubation did not in-crease the As concentration in the soil solution (Fig. 5), consistentwith previous reports of a stronger adsorption of arsenate orarsenite in the acidic pH range (Yamaguchi et al., 2011).

3.5. Plant uptake of As

Herbage (Cut 1) As concentration varied from0.02 to 1.9mg kg�1

with a mean of 0.2 mg kg�1. A preliminary inspection of the datasuggested that As inputs from P fertilisers increased plant As con-centration only during the period when soil As increased rapidly(1888e1947, i.e. when P fertilisers contained large concentrations ofAs), but not in the period after (1959e2011). To analyse the effect ofAs inputs from P fertilisers on plant As uptake using analysis ofvariance, different plotswere grouped into�Pwith orwithout lime,and plant samples from different years in the two periods weretreated as replicates. In both periods, the effect of liming was notsignificant; thus, the data were combined into the�P orþP groups.It is clear that the additions of P fertilisers and the associated Asduring 1886e1947 increased plant As concentration significantly(P< 0.01, on average by 45%), whereas this effect was not significantduring 1959e2011 (Fig. 6A). The estimated total As uptake byherbage varied from 0.02 to 14.2 g ha�1 yr�1 (mean 1.0 g ha�1)among the different plots and years analysed (Fig. 6B). In bothperiods,þP increased plant P concentrations significantly (P< 0.01)and to a similar extent (2.2 fold).

4. Discussion

It is generally thought that phosphate fertilisers are not animportant source of As which might contaminate agricultural soils(Chen et al., 2007; Jiao et al., 2012; Nziguheba and Smolders, 2008).However, results from the present study clearly demonstrate that P

fertilisers applied to the Park Grass Experiment during 1888e1947added a substantial amount of As to the soil, resulting in a neardoubling of the total As concentration in the topsoil. Large con-centrations of As (401e1575 mg kg�1) were measured in thearchived P fertilisers between 1925 and 1947. Such concentrationsare approximately two orders of magnitude larger than those re-ported for P fertilisers worldwide (typically <20 mg kg�1). Thehighest As concentration in P fertilisers reported previously was155 mg kg�1 (Jiao et al., 2012). A review of the concentrations oftrace elements in world resources of phosphate rock (VanKauwenbergh, 1997) showed that the igneous phosphate rock inthe Kiiruna deposit, Sweden, contained anomalously high As con-centrations (77e1300, mean 744 mg kg�1). Therefore, the high Asconcentrations in the P fertilisers used in the Park Grass Experi-ment before 1948 were possible if they were made from such, orsimilar, sources. Similar high-As P fertilisers might have been usedin the UK and elsewhere in the past, although the extent of such useis not known. Since 1948, the source of phosphate rock must havechanged because As concentrations (1.6e20 mg kg�1) droppeddramatically to within the usual range reported worldwide.

The results fromMonte Carlo simulations based onmass balanceagree well with the observed changes in the soil As concentrations,indicating that the main inputs and outputs of As have beenincluded. The As inputs from P fertilisers far outweighed the Asoutputs by plant uptake and potential leaching losses during 1888e1947, and since 1948 the soil As concentrations have remainedstable, suggesting that As inputs and outputs are in balance. Simi-larly, Chen et al. (2007) found minimal changes in soil As concen-tration over a simulation period of 100 years for Californiancropland soils using As input and output parameters that are com-parable to those for the Park Grass Experiment since 1948. Removal

T.N. Hartley et al. / Environmental Pollution 180 (2013) 259e264264

of As by herbage is very small, likely due to limited translocation ofAs from roots to shoots (Zhao et al., 2009). Leaching of As ispotentially larger than plant removal. However, the method used toestimate soluble As (i.e. shaking of 1:1 soil:water suspension for24 h) may have overestimated the As partition into the solutionphase compared with the field conditions. Nevertheless, As solu-bility and potential leaching would increase greatly if the soil isallowed to become anaerobic (Fig. 5). It iswell known that anaerobicconditions induce As mobilization in soil (Masscheleyn et al., 1991;Xu et al., 2008; Yamaguchi et al., 2011). Some possible downwardmovement of As and P in the soil profile was observed in a limedþPplot, consistent with their increased mobility with increasing pH(Tyler and Olsson, 2001; Yamaguchi et al., 2011).

A significant increase in As concentrations in herbage wasobserved in the þP plots compared with �P plots before 1948, butnot after (Fig. 6). A possible reason for this difference is that Asadded with the P fertilisers during the early period has aged in thesoil, resulting in decreased bioavailability (Song et al., 2006).Another reason why a doubling of soil As concentration did notincrease herbage As concentrations more may lie in the competi-tion between arsenate and phosphate, which are chemical ana-logues and are taken up via the same membrane transporters byplant roots (Zhao et al., 2009). Thus, a high availability of P in theþPplots would suppress the uptake of arsenate. None of the herbagesamples analysed exceeded the EU limit for As in feed of 2 mg kg�1

(McGrath and Zhao, 2013). The risk of As entering plants and thefood chain may increase if the soil develops anaerobic conditions,resulting in the reduction of arsenate to arsenite, because the latterdoes not share the phosphate transport pathway (Zhao et al., 2010).

In conclusion, our studyhas revealed ahitherto unreported sourceof As contamination. Although most of the P fertilisers used nowa-days contain low levels of As, historically high-As P fertilisers existedand were used for over 60 years during the early phase of the ParkGrass Experiment, leaving a legacy of As contamination in the P fer-tilised plots. The global reserve of high-grade and low-contaminantphosphate rock available for use as fertiliser is predicted to dwindle(Cordell et al., 2009). Consequently, effective quality control in futurefertiliser manufacturing procedures will be of increasing importanceto ensure As concentrations in P fertilisers remain low.

Acknowledgements

We thank Mr Adrian Crosland for assistance in ICP analysis, MrPaul Poulton for comments on the manuscript and the LawesAgricultural Trust for access to archived plant and soil samples. TheRothamsted Long-term Experiments National Capability is sup-ported by the UK Biotechnology and Biological Research Counciland the Lawes Agricultural Trust.

References

Anon, 2006. Guide to the Classicals and Other Long-term Experiments, Datasets andSample Archive. Lawes Agricultural Trust Co. Ltd., Harpenden, U.K.

Blake, L., Goulding, K.W.T., Mott, C.J.B., Johnston, A.E., 1999. Changes in soil chem-istry accompanying acidification over more than 100 years under woodland andgrass at Rothamsted Experimental Station, UK. European Journal of Soil Science50, 401e412.

Chaney, R.L., 2012. Food safety issues for mineral and organic fertilizers. Advancesin Agronomy 117, 51e116.

Chen, W.P., Chang, A.C., Wu, L.S., 2007. Assessing long-term environmental risks oftrace elements in phosphate fertilizers. Ecotoxicology and Environmental Safety67, 48e58.

Chen, W.P., Li, L.Q., Chang, A.C., Wu, L.S., Chaney, R.L., Smith, R., Ajwa, H., 2009.Characterizing the solidesolution partitioning coefficient and plant uptakefactor of As, Cd, and Pb in California croplands. Agriculture Ecosystems &Environment 129, 212e220.

Cordell, D., Drangert, J.O., White, S., 2009. The story of phosphorus: global foodsecurity and food for thought. Global Environmental Change-Human and PolicyDimensions 19, 292e305.

Garcia, R.A., Rosentrater, K.A., 2008. Concentration of key elements in NorthAmerican meat & bone meal. Biomass & Bioenergy 32, 887e891.

Grant, C.A., Sheppard, S.C., 2008. Fertilizer impacts on cadmium availabilityin agricultural soils and crops. Human and Ecological Risk Assessment 14, 210e228.

Hopkins, D.W., Waite, I.S., McNicol, J.W., Poulton, P.R., Macdonald, A.J.,O’Donnell, A.G., 2009. Soil organic carbon contents in long-term experimentalgrassland plots in the UK (Palace Leas and Park Grass) have not changedconsistently in recent decades. Global Change Biology 15, 1739e1754.

Jiao, W.T., Chen, W.P., Chang, A.C., Page, A.L., 2012. Environmental risks of traceelements associated with long-term phosphate fertilizers applications: a re-view. Environmental Pollution 168, 44e53.

Jones, K.C., Jackson, A., Johnston, A.E., 1992. Evidence for an increase in the cad-mium content of herbage since the 1860s. Environmental Science & Technology26, 834e836.

Jones, K.C., Johnston, A.E., 1991. Significance of atmospheric inputs of lead tograssland at one site in the United Kingdom since 1860. Environmental Science& Technology 25, 1174e1178.

Kohler, I.H., Poulton, P.R., Auerswald, K., Schnyder, H., 2010. Intrinsic water-use ef-ficiency of temperate seminatural grassland has increased since 1857: ananalysis of carbon isotope discrimination of herbage from the Park GrassExperiment. Global Change Biology 16, 1531e1541.

Knights, J.S., Zhao, F.J., Spiro, B., McGrath, S.P., 2000. Long-term effects of land useand fertilizer treatments on sulfur cycling. Journal of Environmental Quality 29,1867e1874.

Masscheleyn, P.H., Delaune, R.D., Patrick, W.H., 1991. Effect of redox potential andpH on arsenic speciation and solubility in a contaminated soil. EnvironmentalScience & Technology 25, 1414e1419.

McGrath, S.P., Cunliffe, C.H., 1985. A simplified method for the extraction of themetals Fe, Zn, Cu, Ni, Cd, Pb, Cr, Co and Mn from soils and sewage sludges.Journal of the Science of Food and Agriculture 36, 794e798.

McGrath, S.P., Zhao, F.J., 2013. Concentrations of metals and metalloids in soils thathave the potential to lead to significant concentrations of contaminants in foodand feed. Soil Use and Management (in press).

McLaughlin, M.J., Parker, D.R., Clarke, J.M., 1999. Metals and micronutrients e foodsafety issues. Field Crops Research 60, 143e163.

Meharg, A.A., Williams, P.N., Adomako, E., Lawgali, Y.Y., Deacon, C., Villada, A.,Cambell, R.C.J., Sun,G., Zhu,Y.G., Feldmann, J., Raab,A., Zhao, F.J., Islam,R.,Hossain, S.,Yanai, J., 2009. Geographical variation in total and inorganic arsenic content ofpolished (white) rice. Environmental Science & Technology 43, 1612e1617.

Meharg, A.A., Zhao, F.J., 2012. Arsenic & Rice. Springer, Dordrecht.National Research Council, 2001. Arsenic in Drinking Water 2001 Update, Wash-

ington, DC.Nicholson, F.A., Jones, K.C., Johnston, A.E., 1994. Effect of phosphate fertilizers and

atmospheric deposition on long-term changes in the cadmium content of soilsand crops. Environmental Science & Technology 28, 2170e2175.

Nziguheba, G., Smolders, E., 2008. Inputs of trace elements in agricultural soils viaphosphate fertilizers in European countries. Science of the Total Environment390, 53e57.

Panaullah, G.M., Alam, T., Hossain, M.B., Loeppert, R.H., Lauren, J.G., Meisner, C.A.,Ahmed, Z.U., Duxbury, J.M., 2009. Arsenic toxicity to rice (Oryza sativa L.) inBangladesh. Plant and Soil 317, 31e39.

Silvertown, J., Poulton, P., Johnston, E., Edwards, G., Heard, M., Biss, P.M., 2006. ThePark Grass Experiment 1856e2006: its contribution to ecology. Journal ofEcology 94, 801e814.

Song, J., Zhao, F.J., McGrath, S.P., Luo, Y.M., 2006. Influence of soil properties andaging on arsenic phytotoxicity. Environmental Toxicology and Chemistry 25,1663e1670.

Stacey, S.P., McLaughlin, M.J., Hettiarachchi, G.M., 2010. Fertilizer-born traceelement contaminants in soils. In: Hooda, P.S. (Ed.), Trace Elements in Soils.Wiley, Chichester, pp. 135e154.

Tyler, G., Olsson, T., 2001. Concentrations of 60 elements in the soil solution asrelated to the soil acidity. European Journal of Soil Science 52, 151e165.

Van Kauwenbergh, S.J., 1997. Cadmium and other minor elements in world re-sources of phosphate rock. The Fertiliser Society Proceedings 400, 1e40.

Van Vuuren, D.P., Bouwman, A.F., Beusen, A.H.W., 2010. Phosphorus demand for the1970e2100 period: a scenario analysis of resource depletion. Global Environ-mental Change-Human and Policy Dimensions 20, 428e439.

Warren, R.G., Johnston, A.E., 1964. The Park Grass Experiment. Report of theRothamsted Experimental Station for 1963, pp. 241e242.

Xu, X.Y., McGrath, S.P., Meharg, A., Zhao, F.J., 2008. Growing rice aerobically mark-edly decreases arsenic accumulation. Environmental Science & Technology 42,5574e5579.

Yamaguchi, N., Nakamura, T., Dong, D., Takahashi, Y., Amachi, S., Makino, T., 2011.Arsenic release from flooded paddy soils is influenced by speciation, Eh, pH, andiron dissolution. Chemosphere 83, 925e932.

Zhao, F., McGrath, S.P., Crosland, A.R., 1994. Comparison of three wet digestionmethods for the determination of plant sulphur by inductively coupled plasmaatomic emission spectroscopy (ICP-AES). Communications in Soil Science andPlant Analysis 25, 407e418.

Zhao, F.J., Ma, J.F., Meharg, A.A., McGrath, S.P., 2009. Arsenic uptake and metabolismin plants. New Phytologist 181, 777e794.

Zhao, F.J., McGrath, S.P., Meharg, A.A., 2010. Arsenic as a food-chain contaminant:mechanisms of plant uptake and metabolism and mitigation strategies. AnnualReview of Plant Biology 61, 535e559.