Managing health risks in urban agriculture: The effect of vegetable washing for reducing exposure to metal contaminants

(cid:129) Household washing only marginally reduces the content of several metals in crops. (cid:129) About 50 % of Pb, Co, Cr and As is removed when urban vegetables are washed. (cid:129) Less than 10 % of Zn, Cd and Ba is removed. (cid:129) The washing effect for different elements correlates with their phytoaccessibility.

• Household washing only marginally reduces the content of several metals in crops.
• About 50 % of Pb, Co, Cr and As is removed when urban vegetables are washed.• Less than 10 % of Zn, Cd and Ba is removed.
• The washing effect for different elements correlates with their phytoaccessibility.

A B S T R A C T A R T I C L E I N F O
Editor: Filip M.G.Tack

Introduction
As the earth's urban population increasessupposedly by >2.5 billion by year 2050 according to the UN (2018)it becomes increasingly important to secure this population's access to healthy food.Urban cropland currently accounts for about 70 million hectares, or 6 % of the total agricultural land on the globe (Thebo et al., 2014).But urban cultivation is gaining momentum, vacant land is being reclaimed for gardening purposes, and an increasing number of households consume urban-grown vegetables (Carlet et al., 2017;Kessler, 2013;Palau-Salvador et al., 2019;Saumel et al., 2012;Warming et al., 2015;Witzling et al., 2011).The significance of urban gardening for building resilience to future food insecurity has been further highlighted in the wake of the Covid pandemic (Lal, 2020;Pulighe and Lupia, 2020;Song et al., 2021).
While urban gardens provide nutritious vegetables they may, however, also be a route of exposure to heavy metals and other soil contaminants (Bretzel and Calderisi, 2006;Clarke et al., 2015;Mitchell et al., 2014;Szolnoki et al., 2013).Studies that have specifically addressed produce from urban cropland have shown higher concentrations of several metals in these than in their supermarket equivalents, and in some instances health-based guideline values have been exceeded (McBride et al., 2014;Saumel et al., 2012;Spliethoff et al., 2016).Since diet, and mainly different vegetables, fruits and grains, is the major exposure pathway for most metals (Cao et al., 2016;EFSA, 2009aEFSA, ,b, 2010;;Glorennec et al., 2016;Parveen et al., 2018), the control of human exposure is often about minimising the contamination of food.So, with an increasing production of food on urban soils comes an increasing need to manage the contamination risks in these environments.
If you search the internet on how to practice urban gardening in a safe manner you get thousands of hits, for example links to highly regarded research institutes and authorities that have formulated advice and recommendations; e.g., The Johns Hopkins Center for a Livable Future (2014) and the Office of Superfund Remediation and Technology Innovation at the US EPA (2011).Besides recommendations that aim at reducing the root absorption of contaminants from the soil, a key advice is to wash the produce before consumption.Astonishingly little has, however, been published on the efficiency of common household washing for the removal of contaminants.The recent review by Bidar et al. (2020), for example, thoroughly presents what was then known about metal contamination of urban vegetables.Among the 45 studies that are summarised in the tables and figures of this review, only one (Folens et al., 2017) analysed the vegetables both before and after washing.Most had washed the vegetables prior to digestion/analysis with the aim to remove adhered soil, but without checking whether it had much effect.There are a handful of other studies that have addressed the issue, but where the analysed vegetables come from local markets or stores (Abdel-Rahman et al., 2018;Al Jassir et al., 2005;Oteef et al., 2015), in which case the pretreatments are unknown as well as whether the location of origin is urban or rural.Among the studies where the researchers have controlled the cultivation and harvest, generalisations are still hard to make, for several reasons: different studies have focused on different vegetables and metals, the washing procedure is not consistent between studies, and the cultivation environments vary.In addition, the data sets found in the published literature are overall small, in most cases based on only a few samples and none hasto the best of our knowledgesought out to characterise and explain the variability seen for different elements.
This paper presents how elemental concentrations in leafy vegetables from an urban environment are affected when the vegetables are washed.The analysis is based on results from a wide multi-element screening approach that targeted 71 elements in paired unwashed and washed samples.We discuss why some elements are significantly removed upon washing while others remain basically unaffected.We further discuss what our results imply in a risk management context, and examine how much individuals who consume relatively high amounts of urban vegetables can reduce their exposure and risk if they wash their produce.

Study site
The vegetables sampled for this study originate from a site called Botildenborg.It is located in southern Sweden, in the city of Malmö -the third largest city in the country, with ca 349,000 inhabitants.Along the inner ring road, trafficked by ca 70,000 vehicles per day, lies Botildenborg as a part of the suburb Rosengård; an area nationally recognised for its high degree of ethnic diversity and socio-economic deprivation.The meeting spot Botildenborg was created here with urban farming and sustainable food production as common denominators in a wider effort to increase integration and foster social, ecological and economic sustainability.

Vegetable cultivation, sampling and washing
Four different kinds of common leafy vegetables were collected during the cultivation season in 2020: lettuce, chard, kale and parsley.The vegetables were cultivated on open land according to Fig. 1, with ca 60-90 m distance to the ring road.Insect nets with a mesh size of 0.8 * 0.8 mm were used to protect the crops from herbivores, insects, snails and other pests.Previous studies indicate that netting has a negligible impact on the vegetables' exposure to airborne particles (van Hengstum et al., 2012).Lettuce, chard and parsley were harvested in late July after approximately 2-3 months of growth, and kale in mid-August after approximately 4 months.In total, 9.7 mm precipitation was recorded in the area during the two weeks preceding the first harvest, but none in the last four days.Before the harvest of kale, 12.5 mm rain was recorded in the previous two weeks, 3.4 mm of which fell in the last four days (data received from Swedish Meteorological and Hydrological Institute, SMHI).
With the purpose of analysing paired samples of washed and unwashed material, leaves from individual plants were carefully divided into two equivalent subsamples directly in the field.The same number of leaves from outer and inner parts of the plants, as well as the same number of larger and smaller leaves, were selected to minimise differences in exposed leaf areas between the two subsamples.To acquire enough material for the chemical analyses, each sample was produced by pooling material from several individual plants that grew close together.In total, 17 samples of lettuce were collected, 16 samples of chard, 15 of kale and 15 of parsley (63 in total); all divided in two subsamples, of which one was prepared for analyses without washing and the other one being washed prior to further preparation and analysis.The samples were put in 5 L plastic bags in the field and kept cool.During the collection of vegetable samples, a small portion of soil was also taken from the rootzone under each plant so that all vegetable samples were paired with a sample of the soil they had grown in.
The washing procedure, carried out within 24 h after sampling, was designed to be as similar as possible to common household washing.Vegetables were placed in a common household plastic colander and rinsed under cold, running tap water for 10 s while mixing the leaves by hand.Disposable, powder-free, latex gloves were used during all sample handling.The colander was lightly shaken over the sink to get rid of excess water and vegetables were blotted dry with tissue paper.All subsamples, both washed and unwashed, were then dried in aluminum foil trays at 60 °C for 36 h, but without mixing washed and unwashed subsamples in the drying cabinet.The dry vegetables were put in the same kind of plastic bags (clean ones for the washed samples) that were used to collect them, as to minimise the contact with new materials, and homogenised directly in these bags.Despite the care taken, it should still be mentioned that there is a risk that some of the particles are lost during drying and homogenisation, and hence, lead to underestimation of the concentrations.For each crop type, the fresh:dry weight mass ratio was determined for 5 subsamples so that the dry weight concentrations could be converted to fresh weight equivalents for comparisons with data from other studies (as in Table 2 + 4 and Section 3.3).These analyses gave an average water content of 87.8 % for all crops; 80.0-83.5 % for parsley, 83.0-85.5 % for kale, 90.7-91.9% for chard and 94.4-94.6 % for lettuce.

Chemical analyses
Approximately 0.50 g of dried and homogenised plant material, in one washed and one unwashed subsample, were sent together with 0.50 g of the paired soil samples for digestion and further analyses at the accredited laboratory ALS Scandinavia in Luleå, Sweden.All samples were treated and analysed according to the procedures outlined below.

Sample digestion
All the laboratory work was carried out in Class 10,000 clean laboratory areas, using plastic utensils that had been thoroughly acid washed, and with all chemicals used of Suprapur grade.The (pseudo)total concentrations of metals in soil samples were determined after Aqua Regia digestion of sieved (<2 mm) and dried (at 60 °C) material, following standard protocols.Hereby, 0.50 g of soil was mixed with 7.5 mL HCl and 2.5 mL HNO 3 , heated to 130 °C for 2 h and diluted to 50.0 mL with Milli-Q water prior to analysis.The digestion procedure chosen for the vegetable samples is one that renders a complete digestion of both plant tissues and attached material such as fine soil particulates or dust.Hereby it enables the determination of total concentrations of the analysed elements.First, 10 mL of concentrated Suprapur grade HNO 3 was added to each sample, whereafter the tubes were loosely cupped and left inside a fume hood for overnight pre-digestion at room temperature.The tubes were then placed on a graphite heating block maintained at 120 °C for an hour.After cooling off, 0.02 mL of 48 % HF was added in order to dissolve remaining siliceous phases.The tubes were tightly closed and placed in a bench-top laboratory shaker for 2 h.The digests were diluted to 20 mL with Milli-Q water and further 50-fold with 1.4 M HNO 3 prior to the ICP analysis, providing a total dilution factor of approximately 1000× (v/m).
In order to assess the repeatability of the analytical procedure, every tenth sample was prepared and analysed in duplicate.Each preparation batch (46 samples) included a minimum of two preparation blanks and two samples of reference materials.The reference materials included two matrix-matched certified reference materials (CRMs) from WEPAL; IPE 776 (lettuce) and IPE 167 (french beans), with certified concentrations available for 21 elements (As, B, Ba, Ca, Cd, Co, Cr, Cu, Fe, Hg, K, Mg, Mn, Mo, Na, Ni, Pb, Se, Sr, V and Zn).Additionally, a sample of spirulina powder that is routinely used as an internal control material at the analysing laboratory was included, selected on the basis of its good homogeneity and measurable concentrations of almost all elements.Concentrations in this internal control material have been established through cross calibration using the CRM NIST1547 (peach leaves).

Quality control and quality assurance (QC/QA) procedures
The accuracy of the ICP-SFMS analyses was evaluated from results obtained for the reference materials.For the elements that were found at concentrations > 10 times the LOD in all samples, the deviation between obtained and tabulated concentrations was in general within 10 % of the relative standard deviation (RSD).
Out of 71 elements measured by the ICP-SFMS, however, only 50 were selected for further evaluation.The reason for excluding 21 elements was that these were frequently found in concentrations close to or below these elements' levels of detection (LOD).The excluded elements were, firstly, ten ultra-trace elements (Au, Br, I, Ir, Os, Pd, Pt, Ru, Ta, Te) that were below LOD in at least 60 % of the analysed samples, and below the level of quantification (LOQ) in all samples.Secondly, another 11 elements were excluded (Ag, Cs, Ge, Hg, Nb, Re, Rh, Se, Sn, Sr and W).These were for the most part detected, but frequently (in >25 % of the samples) below LOQ.The LODs were calculated as three times the standard deviations of the concentrations in preparation blanks, while the LOQs were estimated as ten times the standard deviations of the blanks (N = 21).The vast majority of the reduced set of 50 elements were found in concentrations > LOQ in all samples, and with a good margin.
The methodological reproducibility for the selected 50 elements was assessed from the RSD values of the duplicate samples.The RSDs varied from 5 to 41 %, with a mean RSD of 21 % when considering all elements.For the elements that were present at concentrations >10 times the respective LOD in all samples, the average RSD was 16 %.For the remaining elements, the average RSD was 30 %.

Calculation of washing effect
The washing effect, expressed as the fraction (%) of an element that is removed from a certain sample upon washing, is calculated according to Eq. ( 1), where C uw is the element concentration in the unwashed subsample and C w is the concentration in the parallel, but washed, subsample.Both concentrations are given on a dry weight basis.
It should be noted that this concentration-based calculation will result in a slight underestimation of the washing effect, due to the removal of (some) exogenic material between the analyses of Cuw and Cw.The significance of this should, however, be low relative to other uncertainties, in e.g., the matching of unwashed and washed subsamples and in the chemical analyses.A simple example can help to clarify: Dried and unwashed plant material usually contains <1 % w/w adhering particles (Caille et al., 2005;McBride et al., 2014), in which case 0.50 g of unwashed plant material should have given only 0.495 g had this material been washed and all adhering particles removed.Thus; when 0.50 g was digested of both the unwashed and washed material, the analysed amount of the latter was slightly exaggeratedhowever, only 1 % at the most (0.5 vs. 0.495 g).If the concentration of a certain element was halved after washing, the concentration-based calculation of the washing effect would give a 50 % reduction.A mass-based calculation on the other hand, correcting for the 1 % extra material in the washed sample, would instead render a 50.5 % removal.Optimally, mass-based calculations should have been performed too, but for that we would have needed to know the amount of adhering material.

Statistical analyses
All statistical tests for this study were performed using R (R Core Team, 2017).Initially, a principal component Analysis (PCA) of the washing effect was done to see if clusters of elements could be identified.Results from this analysis are presented in the Supplementary material (Fig. S1) and show a tight clustering of the rare earth elements, REEs (Y, Sc and the lanthanids Ce, Dy, Er, Eu, Gd, Ho, La, Lu, Nd, Pr, Sm, Tb, Tm and Yb).Further, all the individual REEs' display a strong correlation to the total rare earth element concentration, ΣREE (Fig. S2), something to expect given these elements' similar physicochemical characteristics.The concentrations of all REEs were therefore summarised and evaluated as the ΣREE concentrations throughout the rest of the paper, an approach frequently adopted in studies of these elements' environmental distribution (MacMilllan et al., 2017;Squadrone et al., 2019).Throughout the main text below the statistical significance of the washing effect for individual elements is assessed from paired t-tests, assuming the difference between the concentrations in unwashed and washed samples to be approximately normally distributed.The p-values were adjusted with Holms sequentially rejective multiple test procedure.

The representativeness of the Botildenborg data for studying "urban vegetable" characteristics
As outlined already in the introduction, there are plenty of examples in the scientific literature that point to the elevation of several metals (like As, Ba, Cd, Cu, Ni, Pb and Zn) in urban cultivated soils (Bretzel and Calderisi, 2006;Clarke et al., 2015;Mitchell et al., 2014;Szolnoki et al., 2013), even though an increased awareness among both private gardeners (e.g.allotment holders) and commercial actors (like in Botildenborg) probably means that cultivation in highly contaminated areas is unlikely.
Basic soil geochemistry and concentrations of some urban metal contaminants in the soil of Botildenborg are summarised in Table 1.Concentrations of all analysed elements are given in the Supplementary material (Table S1).Even though some metals are elevated in Botildenborg in comparison to normal agricultural land (Table 1), the Botildenborg soil in itself is not sensu stricto contaminated land.Concentrations of metals in the Botildenborg soil are below the national guideline values from the Swedish Environmental Protection Agency, and they are relatively low when compared to concentrations previously reported for other urban soils used for cultivation.The latter is illustrated in Table 1 by the data from Qvarforth

Table 1
Basic soil geochemistry in soil samples from Botildenborg, together with concentrations of some metals that are often acknowledged as urban contaminants.The analysed soil samples were paired with the analysed vegetables (thus, n = 63).For the metal contaminants are also given concentrations reported for other urban soils (Qvarforth et al., 2022), guideline values for "sensitive land use" according to the Swedish Environmental Agency (Swedish EPA, 2009), and the 95th percentile concentrations (P95) in Swedish agricultural soils (Eriksson, 2021).The EPAs guideline values for sensitive land use indicate the maximum permissible concentrations without restricts of e.g., the use of the soil for cultivation purposes.However, even in cases where the soil complies with applicable healthbased guidelines, elevated concentrations of various contaminants may still be found in crops (Murray et al., 2009).One plausible explanation is that some plants will accumulate metals during growth.More widely applicable though is the fact that plants are also likely to receive contaminants after deposition of aerial soil/dust, which may have been resuspended from nearby roads and industrial areas.In a recent paper by Li et al. (2022), it was even suggested that foliar dust may be a better indicator of the contamination burden on a specific site than concentrations of contaminants in the soil.Ensuring a healthy substrate as the crop growth media will not affect the deposition of air-borne particles from near-by areas, and the ideal scenario would be if these particles were efficiently removed by washing of the crops during food processing or preparation.
Table 2 summarises how the concentrations in the Botildenborg vegetables compare to those reported from previous urban studies, mainly from allotment gardens in northern Europe and in the US, and also to concentrations in commercial produce.To match the previously published data, Table 2 focuses on leafy and washed vegetables.It shows that average As, Ba and Cd concentrations in the Botildenborg vegetables were higher than reported for the other urban sites, while Pb and Zn were generally lower.It is thus reasonable to assume that the Botildenborg samples are fairly representative for vegetables grown under conditions found in urbanized regions of e.g., northern Europe and north America.Most importantly, however, the data summarised in Table 2 implies that contaminant metals are generally elevated in urban vegetables compared to such found on the European commercial market; the latter inferred from data from the European and Swedish food safety authorities.Although admittedly not a dramatic increase, even small differences in the crop metal content may significantly affect the total exposure for consumers who rely on localised sources; something which was highlighted as an important conclusion in the study by Augustsson et al. (2018).The question, then, is how much the concentrations are affected by washing.

The washing effect
Table 3 shows a summary of the retrieved element concentrations in unwashed and washed samples, respectively, when all crops are considered.Data on individual crops can be found in the Supplementary material (Table S2).Elements in dark grey are those for which a significant reduction in concentration was seen after washing (p < 0.005).Individual pvalues for all elements are presented in the Supplementary material, Table S3.The overall variability in washing effect is illustrated in Fig. 2a, and for individual crops in the Supplementary material, Fig. S3.The concentrations of most elements in/on unwashed vegetables, and especially those that are significantly reduced when vegetables are washed, were highest in lettuce and lowest in parsleyindicating that these two crops were the ones that captured exogenic particles most and least efficiently in this study.Plant-specific factors that affect the number of particles trapped by above-ground crops are the foliar surface area, the roughness/ waxiness of the leaf surfaces, and the duration of growth (Schreck et al., 2012;Shahid et al., 2017;Xiong et al., 2014).It has previously been found that parsley may contract soil particles (or rather lead, which is an element that is often associated with adhering soil particles) more efficiently than e.g.lettuce (Schreck et al., 2012), but to about the same degree as chard (Osma et al., 2012).The deviation in this study may be due to the variety of parsley investigated here being a broad-leaved one.
Besides the washing effect being different for different elements, and significant for about half of them (as indicated by asterisks above the boxes in Fig. 2a) there is also a great variability observed for individual elements.As a rule, this variability is higher at the right-hand end of Fig. 2a, at which elements that are more efficiently removed by washing are shown.The average washing effect for paired samples ranged from −1.7 %

Table 2
Mean and median concentrations of a selection of urban contaminant elements in Botildenborg (this study), in mg per kg dry weight, compared to concentrations found in urban vegetables in other studies and in commercial vegetables.All concentrations are given for washed vegetables.When concentrations in the original reference were given per kg fresh weight, a conversion into dw was made by assuming a water to dry mass content of 0.88:0.12,which is the average ratio in our analysed crops removal of K to 45 % for Ti (Fig. 2a).The element with the lowest, but still significant, effect was Bi with a mean removal of 18 %.Thus, while significantly lower after washing, some 80 % of the original Bi content still remained.
However, since exactly the same material can't be analysed both before and after washing, despite the care taken to create (as far as possible) matching samples, there were a number of paired samples in our material that displayed negative washing effects (Fig. 2a), i.e., with higher concentrations in the washed subsamples.This phenomenon is also noted in other studies, which have aimed at making the same comparison (Ercilla-Montserrat et al., 2018;Folens et al., 2017;Noh et al., 2019;Pancevski et al., 2014).The majority of those that have previously discussed the reduction in element concentrations after washing have, however, only compared mean concentrations in unwashed and washed material, i.e., without any sample pairing (Kugonic and Kopusar, 2000;Mohite et al., 2016;Osma et al., 2012;Pancevski et al., 2014;Sattar et al., 2013;Singh and Kumar, 2006).Besides reducing the risk of misinterpretations based on pairing of material that is in fact not identical, simply comparing mean concentrations between the two groups is also more robust when assessing long term exposure and health risks.When comparing the mean concentrations in washed and unwashed material (Fig. 2b), the reduction is generally higher than in Fig. 2a.It then ranges from −0.33 % (K) to 68 % (∑REE).

Key controlling factors
Among the 'major vegetable elements', i.e., elements that are presented with concentrations > 0.1 g/kg in Table 3 were, first, major plant nutrients (P, K, Ca, Mg and S) that are efficiently and actively taken up via the roots to accumulate in internal plant tissues.The second group, which partly overlaps with the first, consists of major soil constituents (Si, Al, Fe, Ca, K, Na and Mg).These 7 elements are the most common metals in natural soils, with average oxide concentrations in e.g.European topsoils of 67.7 %, 11.0 %, 3.51 %, 0.92 %, 1.92 %, 0.80 % and 0.77 %, respectively (Salminen, 2005).
Comparing the "plant nutrient" and "major soil constituent" groups we see that all the plant nutrients are basically unaffected by washing, i.e., they are found at the left-hand end of Fig. 2.This applies to Ca and Mg too, despite their concomitant high abundance in geogenic material.The remaining major soil constituents (except Na) shows the opposite response and plot at the right end of the figure, with a significant removal upon washing.This should be due to them being relatively common in attaching particles from cultivated soils and airborne dust.The clear distinguishment between these two groups basically points to the two factors that will control the possibility to wash off a certain element: 1) its susceptibility for root uptake, and 2) its abundance in surface soils and airborne dust.
Fig. 3 visualises how the washing effect for different elements in the Botildenborg material correlates with their availability for plant uptake, here approximated by their average bioconcentration factors (BCFs) in lettuce.The elements' BCFs are given by the ratio of their average concentration in edible vegetable tissues to their average concentration in paired soil (Mirecki et al., 2015).The presented BCFs are based on concentrations in washed vegetables, analysed according to the description in Section 2.3, and pseudo-total concentrations in soil analysed by ICP-MS at the accredited laboratory ALS Scandinavia in Luleå, Sweden, following digestion with Aqua Regia.It is worth repeating that we included hydrofluoric acid in the vegetable digestion protocol (see Methodology).Hence, the extracted concentrations in vegetables, as well as the resulting BCFs, were slightly elevated compared to more standard plant digestion procedures that typically do not include HF.However, the main conclusion to draw from Fig. 3 is that there is a clear correlation between the washing effect and the BCF.The more susceptible an element is to root uptake, the lower the effect of washing on reducing the concentration of that element.As an alternative to showing how an element's susceptibility to washing decreases with its availability for plant uptake (towards the left in Fig. 2), one could instead have shown how the washing potential increases (towards the right in the figure) with the element's correlation to e.g., Al or Ti.These are elements with negligible plant availability, and for which the route of transfer is well known to be primarily via adhering particles (Engström et al., 2008).McBride et al. (2014), for example, showed in a study of urban vegetables in Buffalo and NYC that the Pb concentrations in vegetables were strongly correlated to Al; something which they argued was an indication of vegetable Pb contamination occurring primarily via soil splash and aerial deposition, and not via root uptake.For the Table 3 Minimum, maximum, mean and median concentrations of the analysed elements in unwashed and washed vegetables, without any distinguishment between crop types.All concentrations are given per kilo biomass dry weight (dw).For comparisons with fresh weight (fw) concentrations, the table figures can be multiplied by 0.12 since the average water to dry mass content of our analysed crops were 0.88 to 0.12.The elements for which concentrations are significantly lower in the washed subsamples are marked in the darker grey nuance.
Botildenborg material, the correlation between concentrations of Ti in vegetables and all other elements are shown in the supplement, Fig. S4.The correlation to Ti clearly increases with the washing efficiency, i.e. towards the right in Fig. 2, which points to an increased association with exogenic material.Fig. 4 exemplifies this pattern by showing 4 metals that are often elevated in urban environments; 2 of which are not significantly removed by washing (Cd and Ba), and two which are (Pb and Co).The importance of adhering material may intuitively make one suspect that elements with a high abundance in the soil would respond better to washing than less abundant ones.
To see whether a high occurrence of an element in the soilthus, probably also in adhering particlespromotes an efficient upon washing, Fig. 5 shows the relationship between the washing effect and the aqua-regia-extractable concentrations of the different metals in soil.Fig. 5a focuses on major soil constituents, which appear in concentrations that are often many orders of magnitude higher than for the less abundant elements in Fig. 5b.Despite the former elements' high concentrations, most of them are still reduced by less 10 % when vegetables are washed.Fig. 5b reaffirms the poor relationship between elements' abundance in soil and their removal from vegetables upon washing.In this context we can recall the advice to wash produce before consumption.The rationale behind this common advice is that adhering particles are then removedand assumably so most of them.Since several elements are indeed efficiently reduced, the logic is essentially correct.However, for elements that are efficiently taken up by plantslet's arbitrarily say with BCFs > 1the amount internalised in biogenic plant tissues will by far exceed the amount associated with adhering geogenic material.As mentioned before, no more than a few % of the vegetable dry weight is likely ascribed to adhering particles.It may sound like a significant amount but is negligible for elements with a high plant uptake, where the removal of even all adhering particles will have only a minor impact on the concentration.Let's take Mg, for example.It's BCF is very to 1 (Fig. 3a), meaning that the concentration in Fig. 2. The washing effect, or the % of different elements that is removed from leafy urban vegetables upon washing, illustrated by a) boxplots that show the variability in effect for paired unwashed and washed samples (N = 63).The red asterisks above the boxes mark elements for which the concentration was significantly reduced (p < 0.05).The box is delimited by the lower and upper quartiles of the calculated washing effect, with whiskers extending to show the 5th and 95th percentiles.The median and mean values are indicated by horizontal lines and x.es in the boxes.Subfigure b) shows the mean washing effect when comparing mean concentrations in unwashed and washed vegetables, just treated as two separate groups (i.e., without sample pairing).vegetable and soil is about the same.Consequently, the percentage of Mg removed when a vegetable is washed could never exceed the percentage of adhering particles.
The maximum washing effect can be assessed from e.g. the fraction of removed Ti, since this is an element essentially found only in exogenic material, and also is the element least responsive to washing according to Fig. 2a.The average effect, when all crops are considered, are then about 66 % (Fig. 2b).

Contaminant metals
The element removal upon washing is obviously particularly interesting for potentially toxic contaminant metals.While the degree of contamination by different metals is site specific, the main "risk metals" from a crop contamination perspective are usually considered to be As, Cd, Cr, Cu, Hg, Pb, Zn, Sb, Co and Ni (Tóth et al., 2016).These are primary metals for which elevated concentrations in soils may result in toxicological responses among humans that consume vegetables grown thereon.Mercury was not assessed at Botildenborg because concentrations were frequently <LOD, but the remaining elements are further discussed below.Barium is also discussed since it is addressed in studies that regard metal contamination -even though seldomly posing a threat to human health after vegetable consumption.
As indicated in Table 3 and Fig. 2, the concentrations of Ba, Cd, Zn and Cu were not significantly reduced when the vegetables were washed.For these, average concentrations decreased by only 5, 7, 7 and 13 %, respectively (Fig. 2b).For the elements for which a significant removal was found, the concentrations were reduced in the order: Pb (56 % lost) > Co (56 %) > Cr (55 %) > As (45 %) > Sb (35 %) > Ni 33 %.Under the circumstances that prevail at Botildenborg, and under the evaluated growth season, Pb was thus the contaminant metal that was most efficiently removed.Almost half of the original Pb content did, however, remain even after washing.And for Ni, which was the element with the lowest significant effect, almost 70 % remained (Fig. 2b).
The washing effect according to other studies that focus on leafy crops and can be comparable to ours is summarised in Table 4.As can be seen, the large variability in washing effect found at Botildenborg (Fig. 2) agrees with previously reported findings.It is also in accordance with results presented for other types of crops (by e.g.Abdel-Rahman et al. (2018), Nabulo et al. (2010) and Nabulo et al. (2012)).However, despite the variable effect, the overall picture across these studies is that significant amounts of Pb are generally removed from above-ground plant tissues by washing, but a much lower proportion of e.g.Cd and Zn (Abdel-Rahman et al., 2018;Al Jassir et al., 2005;Ferri et al., 2015;Folens et al., 2017;Kugonic and Grcman, 1999;Nabulo et al., 2010;Noh et al., 2019;Sattar et al., 2013;Žalud et al., 2012).Keeping the above discussion about key factors in mind, these contaminant metals differ in solubility, and hence also in availability for plant uptake.Lead is the most abundant heavy metal in soils, but its solubility and mobility is strongly restricted by its efficient sorption onto clays, organic matter and Fe/Mn (hydr)oxides (Kabata-Pendias, 2011;Salminen, 2005).Thus; the relatively high abundance of Pb in geogenic material in combination with its low phytoavailability suggests that a significant proportion of Pb associated with above-ground leafy tissues should originate from atmospheric deposition and/or soil splash rather than from root uptake; something which has also been suggested before (De Temmerman et al., 2012;Douay et al., 2008;Egendorf et al., 2021;McBride et al., 2014;Nabulo et al., 2012).The two elements, after Pb, for which washing had the greatest effect were Co, Cr and As (Fig. 2), and at higher pH values these too are efficiently retained by sorption to clays, organic matter and various (hydr)oxides (Bissen and Frimmel, 2003;Kabata-Pendias, 2011;Salminen, 2005;Warren et al., 2003).Cadmium and Zn, on the other hand, are more easily mobilised in soils and are more efficiently extracted from the soil pore water by plant roots (Abdu et al., 2011;Chang et al., 2014;De Temmerman et al., 2012;Kabata-Pendias, 2011;Nabulo et al., 2010;Voutsa et al., 1996).Hence the washing effect for the reduction of Cd or Zn on leafy produce is much lower.
As understood from the brief discussion above, the uptake of metals by plants is strongly affected by the natural soil geochemistry, and it differs between plant species and varieties (Bidar et al., 2020;McBride et al., 2014;McLaughlin et al., 2000;Menzies et al., 2007;Nabulo et al., 2010;Waterlot et al., 2013).In addition, foliar adsorption may significantly lead to internalisation of metals derived from airborne deposition (Shahid et al., 2017).On the one hand, one can suspect that the washing effects were higher for many metals at Botildenborg than they may be at many other sites.This could well be the case considering that both soil pH (on average 8.3; Table 1) and the content of soil organic matter (on average 9.1 %) were relatively high at Botildenborg, which should restrict the solubility of most metals.Under a scenario like this, with a low metal solubility and reduced uptake of free ions from the soil solution, the fraction of the various metals associated with adhering material (that can be washed away) should be higher.On the other hand, the washing effect for contaminant metals is probably lower at Botildenborg than at sites where the surrounding contamination is higher, at least for elements with low BCFs.For example, the higher washing effect found for Pb in the study by Egendorf et al. (2022) may partly be explained by the higher soil Pb concentrations in their trials, and to some degree also explained by differences in their experimental approach to the washing procedure.

Risk perspectives
The large, and systematic, differences between metals when it comes to their susceptibility to removal by washing have clear implications for risk management.This aspect is discussed in the following section using Pb and Cd as examples.These two elements represent the metal contaminants most and least affected by washing, respectively; they are both metals for which concentrations in vegetables relatively often result in exposure levels in humans above toxicological reference values; and they are the only two metals with maximum permissible concentrations in fresh leafy vegetables stipulated in the EU Commission Regulation (EC) No 1881/2006; 0.30 and 0.20 mg/kg fresh weight, respectively.Obviously, understanding how Fig. 5.The average washing effect (% removal) for a) major and b) non-major elements as a function of their average log concentration in the soil (μg/kg dw).Elements marked in red are the contaminant metals exemplified in Table 1 + Cr, Co, Mo, V and Sb.crops are contaminated by these two metals are essential for the safeguarding of human health, where such understanding affects which exposure mitigation practices should be recommended gardeners.
As discussed above, accumulating evidence suggests that Pb contamination of vegetables occurs mainly with adhering soil particlesthe significance of which should logically increase with the pollution load of the soil.The contamination by Cd, on the other hand, can be explained mainly by vascular uptake via the roots.The implications of this difference in mode of contamination are illustrated in Fig. 6 which summarises the average concentrations expected to be found in a) commercial leafy vegetables, b) urban leafy vegetables, and c) leafy vegetables grown near more heavily contaminated point sources.Average concentrations are chosen since these are the most relevant for long-term dietary intake, i.e., they relate to the modelling of chronic exposures evaluated using oral reference doses.These concentrations are in the left-hand columns of the figure given for unwashed and washed produce, respectively, with the difference between the two given by the average washing effect from Fig. 2b.Concentrations that exceed the Pb and Cd guidelines for commercial foodstuff are marked in red.The urban vegetable data is retrieved from the same studies as summarised in Table 3, from midsized to large cities in northern Europe and in the US, but with the difference that the data in Fig. 6 is given for fresh vegetables to aid comparison with food guidelines.Concentrations in vegetables near point sources derive from a literature search, where studies were selected if the contamination originated from a distinct point source, but not from e.g., sewage sludge amendment or irrigation with sewage water or industrial effluents.The studies hereby compiled admittedly Table 4 Reduction of metal concentrations in leafy vegetables by washing according to other studies.The reduction is expressed in % of initial concentration lost, as calculated from Eq. ( 1).Negative values are indicative of higher concentration of metal in the washed sample (see Eq. ( 1)).The original references have often included other crops too, and in some cases also control plants from indoor cultivation.Data from not comparable types of crops and from indoor controls are excluded here.The references are arranged based on analysed number of samples (n), from the highest to the lowest.represent only a handful of the total available material (Augustsson et al., 2015;Cui et al., 2004;Dziubanek et al., 2015;Obiora et al., 2019;Pelfrêne al., 2019;Xu et al., 2013;Zhang et al., 2019;Zheng et al., 2007).There is also no clear-cut difference between e.g., "urban areas", "industrialised areas", and "polluted areas".The intention here is simply to give an impression of the differences between the three groups.
For commercial vegetables, shown at the top of Fig. 6, average concentrations of both Pb and Cd comply with existing regulations.Many of the commercial vegetables have been cleaned before delivery to the stores, so whether the consumers make the conscious choice to wash them or not is of less importance.However, vegetables from urban gardens -and especially from land near contaminating point sourcesoften contain both Pb and Cd above the permissible concentration, and washing is then essential for Pb.For Cd it will have little effect.
To make inferences about the possible health risk implications, Fig. 6 also shows the average daily dose (ADD) of Pb and Cd for average and high vegetable consumers.For the latter group, the reasonable maximum exposure (RME) is stated.The ADD is calculated by multiplying the average metal concentration in leafy vegetables from the 3 kinds of areas with the average daily vegetable consumption rate, stated per kg bodyweight.In our calculations, we used information on vegetable consumption from the US EPA exposure factors handbook (US EPA, 2011), where average consumers are reported to eat 2.55 g of leafy vegetables per kg and day, and high consumers 6.0 g/kg/day.Values of ADD depicted in red in Fig. 6 indicate exceedance of toxicological reference values for dietary exposure.For Pb, the exposure is consistently about 2.30 times higher if you eat unwashed vegetables compared to if you wash them.The corresponding figures for the other contaminant metals with significantly higher concentrations in unwashed vegetables were 2.26 (Co), 2.21 (Cr), 1.82 (As), 1.55 (Sb) and 1.50 (Ni).With a difference of this magnitude the washing will in many cases be the difference between an exposure below vs. above the tolerable intake.The advice to wash the vegetables is, for elements like Pb, therefore highly effective at reducing exposure and any associated health risks.Mitigation should also focus on measures that can reduce the deposition of soil splash and airborne dust onto the vegetables in the first place.For example, creating barriers between the crops and nearby roads, or covering the soil with a layer of mulch (Egendorf et al., 2021;Saumel et al., 2012).
For Cd, on the other hand, the difference is negligiblethe ADD is only 7 % higher if the vegetables are consumed without first being washed.For Ba, Zn and Cu the increase is about 5 %, 8 % and 16 %, respectively.In areas with a strong impact from Cd, or other elements that are mainly internalised in plant tissues, it is much more important to focus advice and recommendations on reducing plant uptake; both strategies that can reduce the total concentration of the element in the soil, and strategies that lowers its solubility and phytoavailability.
In urban areas, a combination of measures is preferable, given that the pollution here is often complex and involves multiple elements.Near point sources, it is more common to encounter a situation where a more restricted number of contaminant metals are present.Under such conditions mitigations could be specifically designed considering how the contaminants present are transferred to above-ground crops.

Conclusions
This study shows how the effect of common household washing varies significantly between elements.For major plant nutrients such as K, P, Ca and Mg, for example, the concentrations are basically unaffected.At the other end of the scale are Fe, Al, Ti and the rare earth elements, for which some 60-70 % are removed.Focusing on traditional metal contaminants, Pb was the element most efficiently removed (on average 56 % lost).But still, about half of the element remained after washing.The effect then decreased according to: Co (56 % lost) > Cr (55 %) > As (45 %) > Sb (35 %) > Ni (33 %) > Cu (13 %) > Zn (7 %) > Cd (7 %) > Ba (5 %).For Cu, Zn, Cd and Ba the concentrations before and after washing were not statistically different.
A clear negative correlation between the washing effect and the different elements' bioconcentration factors (BCFs; their concentration in Fig. 6.Average concentration in fresh (mainly leafy) vegetables, given for fresh weight material, with or without washing, and resulting average daily doses (ADDs) given as the average and reasonable maximum exposure (RME).Maximum permissible concentrations in leafy vegetables in Europe are 0.30 mg/kg for Pb and 0.20 mg/kg for Cd (EU Commission Regulation (EC) No 1881/2006).Vegetable concentrations above these are marked in red.Tolerable daily intake (TDI) levels used for Pb* and Cd are 0.50 and 0.36 μg/kg/day, respectively (EFSA, 2009a(EFSA, , 2010)), with the critical health effects being developmental neurotoxicity in children (Pb) and kidney toxicity (Cd).* Regarding the tolerable intake of Pb, there is actually no currently applicable TDI value due to the difficulties in establishing a threshold dose for non-cancer effects for this element.The fig. 0.50 μg/kg/day represents the lowest BMDL01 value reported in EFSAs latest report on dietary Pb exposure.
vegetables relative to the concentration in soil) shows that the differences can be sufficiently explained by the availability of elements for plant uptake, where elements with a phytoaccessibility (such as Cd) are far less affected by washing than are elements of low phytoaccessibility (such as Pb), for which the transfer to vegetables should occur mainly via soil splash or airborne deposition.Of less importance for the washing effect seen for a certain element is the concentration of the element in the soil.
Since routine washing of vegetables before consumption is a simple and reasonable measure, sticking to this procedure is certainly recommendable under all circumstances.But for risk management practitioners it is still useful to know how efficient it can be, and to what degree it can affect human exposure.In terms of human exposure, consumption of unwashed vegetables increases the average daily intake by 130 % (Pb), 126 % (Co), 121 % (Cr), 82 % (As), 55 % (Sb), 50 % (Ni), 16 % (Cu), 8 % (Zn), 7 % (Cd) and 5 % (Ba).The advice to wash vegetables is therefore, for potentially toxic elements with a low phytoaccessibility (like Pb), highly motivated for reducing exposure and health risks.For toxic elements that are easily taken up via the roots (like Cd), on the other hand, the average daily intake is only marginally affected by washing, and advice should rather focus on reducing the load in the soil itself.

Fig. 1 .
Fig. 1.Map showing the location of the study site in southern Sweden, and an aerial photo of the area under cultivation with distances between different crops and the large road in the upper left corner of the photo.
. The data presented by McBride et al. (2014) originates from17 community gardens and urban farms in NYC and Buffalo, NY.That from Saumel et al. (2012) comes from 28 horticultural plantings within Berlin's centre.Warming et al. (2015) collected their material from 3 gardens in the inner city of Copenhagen, and Folens et al. from 35 private gardens in Ghent.

Fig. 3 .
Fig. 3. Average washing effect (% removal) for lettuce as a function of elements' average BCFs calculated from paired soil and vegetable samples.Subfigure a) shows the major elements according to Table 2, and b + c) the non-major elements.The data points above Mn in subfigure b) relate to elements with BCFs >0.1, which are in focus in subfigure c).Elements marked in red are the contaminant metals exemplified inTable 1 + Cr, Co, Mo, V and Sb.
Mohite et al. (2016)collected from farmers' fields in peri urban Faisalabad, Pakistan, with known metal contamination of soils and crops.Soaking in tap water for 10 min.Results presented only as a mean concentrations in washed and unwashed subsamples (Table1+ 2 in the reference).The washing effects here calculated from these means.Osma et al. (2012): Samples collected from five areas in Istanbul, Turkey (stream side, suburban, industrial, inner city, roadside) and a control village.A composite sample from each area was collected (thus; n = 6) and divided in two.One part was left unwashed and the other "thoroughly washed several times with tap water and thereafter distilled water".The washing effects here are based on data in Table3+ 5 of the reference.Samples from farmers' fields in 5 peri urban areas around Delhi, India, near multiple metal contamination sources.A composite sample from each area was collected (thus n = 5), of which "a part was washed thoroughly with tap water".Results presented only as mean concentrations in washed and unwashed samples (Table6in reference).The washing effects here are calculated from these mean concentrations.Pb\ \Zn smelter plant in Veles, Macedonia (thus; n = 3).Each sample was divided into one subsample that was analysed without washing and one that was washed.The washing procedure not further specified.The data here represents the mean washing effect calculated from the data in Table4of the original publication.Spain (peri urban rooftop and urban courtyard).One composite sample from each site was collected (thus n = 2), and divided into one subsample that was analysed without washing and one that was "rinsed with tap water".The data here represents the mean washing effect calculated from the data in Table5of the original publication.Mohite et al. (2016): Vegetable samples were collected from a farmland in Navi Mumbai (formerly New Bombay), India, located near numerous metal emitting industries.From the publication it appears as if only one composite sample was analysed per vegetable (thus; n = 1).The material was divided in two, where one subsample remained untreated and the other one "was thoroughly washed several times with tap water followed by distilled water" before analyses.