LI Li-jun, LI Kun, JIANG Bao, LI Ju-mei, MA Yi-bing

1 Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing 100081, P.R.China

2 Sichuan Academy of Forestry, Sichuan 610081, P.R.China

3 National Urban Environmental Pollution Control Engineering Research Center, Beijing Municipal Research Institute of Environmental Protection, Beijing 100037, P.R.China

4 National Observation and Research Station of Coastal Ecological Environments in Macao/Macao Environmental Research Institute, Macau University of Science and Technology, Macao SAR 999078, P.R.China

Abstract Determining the appropriate soil cadmium (Cd) criteria for vegetable production is important for ensuring that the Cd concentrations of the vegetables meet food safety standards.The soil extractable Cd criteria for vegetable production are also essential for both food safety and environmental management, especially in areas with a high natural background level.In the present study, soil total and extractable Cd criteria were derived using the approach of species sensitivity distribution integrated with soil aging and bioavailability as affected by soil properties.A dataset of 90 vegetable species planted in different soils was compiled by screening the published in literature in five bibliographic databases using designated search strings.The empirical soil-plant transfer model was applied to normalize the bioaccumulation data.After normalization, the intra-species variability was reduced by 18.3 to 84.4%.The soil Cd concentration that would protect 95% (HC5) of the species was estimated by species sensitivity distribution curves that were fitted by the Burr III function.The soil Cd criteria derived from the added approach for risk assessment were proposed as continuous criteria based on a combination of organic carbon and pH in the soil.Criteria for total Cd and EDTA-extractable Cd in the soil ranged from 0.23 to 0.61 mg kg-1 and from 0.09 to 0.25 mg kg-1, respectively.Field experimental data were used to validate the applicability and validity of these criteria.Most of the predicted HC5 values in the field experimental sites were below the 1:1 line.These results provide a scientific basis for soil Cd criteria for vegetable production that will ensure food safety.

Keywords: Cd, vegetables, soil criteria, species sensitivity distribution, soil extractable Cd criteria

1.Introduction

Soil heavy metal pollution has been regarded as one of the greatest threats to environmental quality (Sarwaret al.2017).Among the heavy metals, Cd is not only extremely poisonous, but it also can be efficiently absorbed and accumulated by organisms (Luczynskaet al.2019).The consumption of vegetables is one of the main ways for heavy metals to enter the food chain (Dinget al.2013),so establishing accurate soil Cd criteria for vegetable production is crucial to ensure food safety.

Previous studies have focused on the derivation of soil Cd criteria for leafy vegetables and root vegetables(Dinget al.2016).Nevertheless, due to the relatively small variety of vegetables examined in the past, the derived soil Cd criteria are still somewhat uncertain.With the enrichment of research data, soil Cd criteria can be derived from the data for a much larger number of vegetable species.Also, due to the differential accumulation of Cd among vegetable varieties, soil Cd criteria for vegetables vary among different vegetable species (Dinget al.2016; Xiaoet al.2018).For example,the soil Cd criteria for fruit vegetables was 1.2 mg kg-1in northern Shaanxi (Chenet al.2021), while the criteria for amaranth, celery, and pak choi were 0.26, 0.34, and 1.94 mg kg-1, respectively (Liet al.2019).Therefore,in the present study, the soil Cd criteria were derived using as many vegetables as possible and integrating bioavailability and soil aging in order to reduce the uncertainty of the derived soil Cd criteria.

The current Cd pollution control criteria and evaluation methods for soil are generally based on total concentrations, which cannot reflect the bioavailability of the Cd in soil (Daiet al.2017).Therefore, assessing the actual impact of Cd on the soil ecosystems by the total concentration alone is insufficient (Kimet al.2015).In addition, because of the complex origin of soil environmental background levels of Cd in China, the soil environmental background levels are quite variable (Zhaoet al.2015).Safe vegetables may come from soils with high background levels, while some soils below the Cd limits could produce unsafe vegetables (Smolderset al.2009; Luet al.2021).It is well known that the uptake of metals by plants correlates with the bioavailable concentrations of those metals (Menget al.2021).Accordingly, the phytoavailability criteria were derived to facilitate the risk assessment of heavy metals in high natural background areas.

The commonly preferred approach for deriving soil quality standards (SQS) or criteria is to use the species sensitivity distribution (SSD) (Dinget al.2016).The SSD analyses can be used to compare the cumulative effects of environmental contaminants on different species,and derive criteria that protect the species diversity.It is a statistical distribution calculated from a sample of bioavailability data and visualized as a cumulative distribution function.When the cumulative probability in an SSD curve reaches 0.05, the corresponding concentration is the HC5value, i.e., that which protects 95% of the species in natural ecological communities (Heet al.2019).Currently, SSD has been applied in several studies that address the water, soil, sediment, and atmospheric environments (Azevedoet al.2015).In Canada, SSDs are recommended in the Canadian Water Quality Guidelines for the Protection of Aquatic Life (Signoreet al.2016).The use of SSDs is also approved within the guidelines of several regulations in the European Union.Although the method of SSD has made great achievements in ecological criteria, few have used soil Cd criteria forecasting for safe agricultural production.Therefore,this study aimed to derive soil Cd thresholds for the safe production of vegetables by using the SSD method.

Some studies have shown that the soil properties and the concentration of Cd in soil may affect the uptake and accumulation of Cd by vegetables (McLaughlinet al.2011; Liet al.2014; Songet al.2015).Because soil pH greatly affects the bioavailability of Cd in soils, soil pH is considered as one of the most important factors, although soil organic matter also affects Cd phyto-availability through Cd immobilization mediated by the association of Cd with particulate organic matter (Kulsumet al.2023).However, the current agricultural soil criteria ignore these factors, so those criteria may not completely guarantee that edible vegetable production meets the food quality standard (Dinget al.2016).Additionally, since most of the Cd concentration data for vegetables and soils have been obtained from artificially spiked soils in the laboratory, a correction factor is used for the effect of time on metal bioavailability (i.e., the aging effect) to minimize the discrepancy caused by the differences in laboratory and field conditions.Therefore, the purpose of the present study is to determine soil Cd criteria by investigating the specific effects of these factors on the soil in China.

In summary, the aim of this study is to derive soil Cd criteria for the safe production of vegetables using an SSD method integrated with soil aging and bioavailability factors as affected by soil properties.The soil Cd criteria were validated by field soils and vegetables.The continuous or scenario soil criteria are provided based on the added risk approach.Also, the soil phytoavailability Cd criteria for vegetables were derived using the SSD method.These results can provide a scientific basis for soil environmental risk assessment and provide guidance for improving soil environmental standards in China.

2.Materials and methods

The procedures published in a previous paper (Liet al.2021) were followed in deriving the soil Cd criteria for agricultural product safety.The derivation of phytoavailability-based criteria followed five procedures,which are briefly described below: (1) Transform the total bioconcentration factor (BCFtotal) into the extractable bioconcentration factor (BCFEDTA); (2) normalize BCFEDTAvalues with aging factors to obtain aged BCFEDTAvalues;(3) correct the aged BCFEDTAvalues with bioavailability models; (4) derive HC5based on EDTA extractable Cd in soil (HC5EDTA) values from SSD curves with correction values; and (5) validate the predictive model for HC5EDTAusing field experimental data.

2.1.Data collection

The method used to build the datasets was based on the Collaboration for Environmental Evidence (CEE)systematic review guidelines (CEE 2013).The data were obtained from literatures published on Science Direct, Web of Science, and the Chinese Core Journal databases (CNKI, Weipu, and Wanfang databases), by using the keywords “Cd”, “soil properties”, “agriculture soil”, and “vegetables”.Finally, the bibliographies of the 73 retrieved articles were consulted.To screen the data,studies must meet the following criteria: (1) experiments conducted on vegetable soils of China; (2) the data of total bioconcentration factor (BCFtotal) in the soil and vegetables were sampled simultaneously; (3) the soil physical and chemical properties were reported in the study; and (4)the aging time was reported in the study.

A total of 445 records were filtered from the 73 articles.There were 65 overlapping records in the datasets.After merging duplicate data (n=65), the remaining 380 records were screened at the full-text level.The application of the inclusion criteria at the full-text level resulted in 254 relevant records.The main reasons for exclusion at this stage were the lack of primary data or aging time, and unclear vegetable species.After merging duplicate control treatment records, a total of 190 records were extracted and they formed the basis of the present study.The establishment of the database is described in Appendix A.

2.2.Data processing and normalization

To eliminate the potential effects of Cd accumulation on vegetables caused by the Cd background concentration in the soil, the added bioconcentration factors (BCFadd-total)was used (eq.(1)):

whereCadd-edibleis the Cd concentration in the edible part of the vegetables increased by added Cd to soil (mg kg-1), andCadd-soilis the added Cd concentration in the tested soil (mg kg-1),Cedibleis the Cd concentration in the edible part of the vegetables (mg kg-1),Csoilis the total Cd concentration in the tested soil (mg kg-1),Cedible-ckis the Cd concentration in the edible part of the vegetables grown in the control soil (mg kg-1), andCsoil-ckis the Cd concentration in the control soil (mg kg-1).

In addition, the added extractable bioconcentration factor (BCFadd-EDTA) was used (eq.(2)):

whereCadd-EDTAis the concentration of added EDTAextractable Cd (mg kg-1),CEDTAis the concentration of EDTA-extractable Cd in the soil (mg kg-1),CEDTA-ckis Cd concentration in the control soil (mg kg-1).

TheCadd-EDTAvalue was calculated by a linear regression model established by using the relationship between theCadd-soilandCadd-EDTA.The concentration of added total Cd and added EDTA-extractable Cd were determined from laboratory experiments.A solution of 0.05 mol L-1EDTA was used as an extractant to assess the extractability of added Cd in 21 Chinese soils.Cadd-soilandCadd-EDTAwere measured through this experiment, and the data were used to build the linear regression model.The concentrations of total Cd and EDTA-extractable Cd added in the soils are described in greater detail in Appendix B.

The concentration of soil EDTA-extractable Cd induced by added Cd (Cadd-EDTA) was calculated using eq.(3):

When the Cd concentration data of the vegetables and soil were obtained from artificially spiked soils in laboratory experiments, a correction should be made for the effect of time on metal bioavailability (i.e., the aging effect) in order to minimize the discrepancy caused by the differences in laboratory and field conditions.The following formula was used to simulate the aging process because of its goodness-of-fit for the aging data of Cd(Croutet al.2006).

whereEtis the lability of added Cd in the soil (mg kg-1),C0is the soil Cd concentration at the initial time (mg kg-1), andtis aging time (d) (coefficient of determination:R2=0.91,P<0.01).Soil pH was determined using soil slurries on the basis of 220% of the field capacity of each individual soil.

Two empirical soil-plant transfer models were obtained from Dinget al.(2013) and Yeet al.(2014).Because of the higherR2values, eq.(5) was applied to normalize the bioaccumulation data.

where BCFtotalis the bioconcentration factor and OC is organic carbon (g kg-1).

The process of normalization for the BCFEDTAvalues was similar as that of the BCFtotalvalues.The relationship between theCadd-soilandCadd-EDTAwas applied to convert the empirical soil-plant transfer model based on total Cd concentration into the model with the EDTA-extractable Cd concentration.Then the empirical soil-plant transfer model based on EDTA-extractable Cd was used to normalize the bioaccumulation data.

2.3.SSD curve construction

whereb,c, andkare the parameters of the Burr III distribution function.

2.4.Soil Cd criteria calculation

The soil Cd concentrations that would protect 95% of the species (HC5) were calculated using BurrliOZ software provided by the Commonwealth Scientific and Industrial Research Organization, Australia (Liet al.2018).

The soil thresholds of Cd corresponding to the 5%cumulative probability on the distribution function curve were calculated using eq.(8) (Campbellet al.2000):

where HC(q) is the soil Cd concentration (in mg kg-1) that can ensure the food security for 95% of the vegetable species in the SSD analysis whenqis 0.05.

The procedure for calculating the soil Cd criteria is similar to the predicted no effect concentration (PNEC)which is widely applied in the derivation of ecological criteria.According to the assessment factor (AF) method provided by the European Union, PNEC or soil Cd criteria were calculated by eq.(9):

where AF is an assessment factor between 1 and 5 which reflects the uncertainty of the data (Gredeljet al.2018).In this study, the value of AF was set as 1 because the soil criteria were derived by the SSD method integrated with soil aging and bioavailability models as affected by various soil properties, and were validated by field data.

So the son said to him, Dear father, you are so poor that I am only a burden to you; I would rather go out into the world and see if I can earn my own living

Because the added risk approach was adopted in this study, the background concentration of soil Cd must be added to calculate the soil total Cd criteria.The soil total Cd criteria can be calculated using eq.(10):

whereCbis the background concentration of soil Cd (mg kg-1).

3.Results

3.1.Normalization of the bioaccumulation data

A total of 90 pairs of soil-plant samples from vegetable production fields across China were selected from the literature.They represented five vegetable species including leafy vegetables, root vegetables, gourd vegetables, legume vegetables, and solanaceous vegetables.The statistical analysis of the data is given in Appendix C.The data were normalized for aging by eqs.(1) and (4), then the empirical soil-plant transfer model was used to normalize the bioaccumulation data.The intra-species variability of the different vegetable species is shown in Appendix D.Taking the soil condition of pH 6.5 and OC 20 g kg-1as an example, the intraspecies variability was reduced by 18.3 to 84.4%after the normalization (Appendix E).The strongest decline occurred in ‘Qing youcai’ belonging to the leafy vegetables, and the coefficient of variation (CV) dropped from 0.58 to 0.091.The reductions in the intra-species variability values suggested that the normalization removed the effects of soil properties to a certain extent.

3.2.Construction of SSD curves and derivation of the soil total Cd criteria

The SSD curves were constructed with the Burr III distribution based on the normalized BCF data for the 90 vegetable species.The fitting parameters based on soil properties are shown in Appendix F.Taking pH 6.5 and OC 20 mg kg-1as an example, the SSD curve was S-shaped.The order of Cd absorption capacity among the vegetable species is shown in Fig.1.The SSD curves indicated that the cumulative frequency distributions of various vegetable species were different.Most of the gourd vegetables fell in the bottom of the SSD curve, while the root vegetables were almost in the middle of the SSD curve.For the SSD curve,capsicum was the strongest Cd accumulator among the vegetable species.The cumulative frequency of capsicum was 1.1%.Due to the different distributions of the various vegetable species in the SSD curve, an accurate HC5could not be derived by a single vegetable species database.For example, in the present study,the cumulative frequencies of the gourd vegetables(e.g., cucumber) and leafy vegetables (e.g., edible amaranth) both reached about 5% in the SSD curve.Furthermore, different sub-species from the same vegetable species also lead to different Cd accumulation values.For example, the cumulative frequencies of‘Gailiangdongbai’ and ‘Beijingxiaoza 60’, which are both Chinese cabbage, varied greatly.For this reason,more vegetable species are needed to derive the soil Cd criteria to ensure their accuracy.In addition, the soil Cd concentration (HC5total-add) was 0.30 mg kg-1with the condition of pH 6.5 and OC 20 mg kg-1according to the SSD curve.

Fig.1 The distribution of Cd uptake capacity among the 90 vegetable species negatively correlated with soil Cd concentration.The soil Cd concentration for each species under different soil conditions was calculated using the corresponding bioconcentration factor value based on the National Food Safety Standard (GB 2762-2017 2017) for Cd in vegetables and normalized to soil pH 6.5 and OC 20 g kg-1.The detailed list of species names and their corresponding analysis data are given in Appendix G.

Previous studies had shown that soil properties have a strong effect on the transfer of Cd from the soil to plants.Therefore, variations in soil properties should be considered when establishing soil Cd criteria for vegetables.In this study, the effects of soil properties on derived HC5total-addare visualized in Fig.2.The details of parameters for the SSD curves based on soil pH and OC are shown in Appendix F.From Fig.2-A, the curve was the steepest at pH 5 and the flattest at pH 8.The variance of Cd accumulation was lower in the acidic soils than in alkaline soils, which was probably due to differences in the ability to activate Cd around the rhizosphere soil by vegetable roots in alkaline soils.With an increase in soil pH, the HC5total-addvalues increased.The distance among the SSD curves showed different increases from thex-axis for each increment of one pH unit, probably due to the influence of soil pH on Cd speciation in the soil solution (Dinget al.2018), and the Cd concentration also showed different increases for every pH unit increase.Other cumulative frequencies also showed similar trends, indicating a stronger Cd buffering capacity in alkaline soils.In addition to the soil pH curves,similar variation trends were observable in the OC curves.The flattest curve was the OC-30 curve (Fig.2-B).With an increase in the OC content, the HC5addvalue increased.In summary, the soil Cd thresholds for vegetables should be determined by considering the combination of soil pH and OC.

The HC5values were numerically derived from each SSD curve, and they ranged from 0.10 to 0.48 mg kg-1,with a 4.8-fold variation among the five scenarios.Finally,the predictive model for HC5addwas then built using multiple regressions on the two soil properties (pH and OC).The equation is:

where HC5total-addis the soil Cd concentration to protect 95% of the cultivars (mg kg-1) and OC is the organic carbon (g kg-1).

The calculation based on added Cd is shown as the continuous criteria in Table 1.Moreover, Table 1 also shows the scenario criteria for different combinations of soil pH and soil OC.The added soil total Cd criteria varied from 0.10 to 0.48 mg kg-1in soil with a pH from 5.0 to 7.0 and OC from 10 to 30 g kg-1.The soil total Cd criteria were calculated using the natural background value of 0.13 mg kg-1and also compared with the current Chinese soil quality standard.This comparison indicated that the derived soil total Cd criteria were slightly lower than the current soil quality standard when the soil pH is less than 5.5.

Table 1 Soil Cd criteria for agricultural product safety and the current soil quality criteria in China

3.3.Validation of soil total Cd criteria for vegetables

The predictive model of HC5total-addwas applied to the field experiment data to test its validity and applicability (these data are summarized in Appendix I).The relationship between the predicted HC5total-addvalues and the field experiment Cd concentration is shown in Fig.3.

Fig.3 Comparison of predicted HC5total-add values and the field soil Cd concentrations from different field sites.HC5total-add, the soil total Cd concentration for protecting 95% of the species.

The soil Cd concentrations in the field ranged from 0.15 to 1.54 mg kg-1, with an average value of 0.60 mg kg-1.The predicted HC5total-addvalues ranged from 0.14 to 0.71 mg kg-1, with an average value of 0.40 mg kg-1.Most of the predicted values were within the 95% confidence intervals, demonstrating the applicability of the predictive models.Most of the predicted HC5total-addvalues in the 14 field experimental sites were below the 1:1 line, although two of the 14 field experimental sites were above the 1:1 line.Most of the predicted HC5total-addvalues were less than the field experiment Cd concentrations.This difference indicated that the predicted soil Cd criteria could protect the vegetables from exceeding the national food safety standard.Therefore, the established model was reliable in predicting the soil Cd criteria.

3.4.Derivation of soil EDTA-extractable Cd criteria

The SSD curves fitted with the Burr III function based on the normalized BCFEDTAdata of the 90 vegetable varieties are shown in Fig.4.Taking the condition of pH 6.5 and OC 20 mg kg-1as an example, the SSD curves clearly show that different vegetable species vary with respect to soil Cd accumulation.The cumulative frequency distribution of the soil EDTA-extractable Cd for different vegetable species was similar to the cumulative frequency distribution of soil total Cd.In this study, solanaceous vegetables were distributed at the bottom of the curve,which indicated that they were high Cd accumulators,while some leafy vegetables were low Cd accumulators.In addition, all kinds of root vegetables were in the low Cd accumulator category.Moreover, the HC5EDTA-addvalue was determined as 0.08 mg kg-1with the condition of pH 6.5 and OC 20 mg kg-1.

Fig.4 The distribution of Cd absorption capacity values of the 90 vegetable species, which are negatively related to the soil EDTA-extractable Cd concentration.The soil EDTA-extractable Cd concentration for each species under different soil conditions was calculated using the corresponding bioconcentration factor value and normalized to soil pH 6.5 and OC 20 g kg-1.The detailed list of species names and their corresponding analysis data are given in Appendix J.

The HC5EDTA-addvalues were strongly affected by soil properties.The effects of soil properties (pH and OC)on derived HC5EDTA-addare shown in Fig.5.The variation trends of HC5EDTA-addwith different soil properties were similar to those of HC5total-add.As the pH increased from 5 to 8, the SSD curves shifted to the right along thex-axis, suggesting that the values of HC5EDTA-addincrease with increasing soil pH.In alkaline soils, soil Cd could be bound with carbonate, and partially co-precipitated with anions such as OH, resulting in a reduction in the bioavailability of Cd.Therefore, the low accumulation of Cd in vegetables led to high HC5EDTA-addvalues.In addition, the HC5EDTA-addvalues also increased with the increasing OC concentration.The details of parameters for the SSD curves based on soil pH and OC are shown in Appendix F.

Fig.5 The curves of vegetable species sensitivity distribution (SSD) at different soil pH and organic carbon (OC) levels.A, OC of 20 g kg-1.B, soil pH 6.5.The soil EDTA-extractable Cd concentration for each species under different soil conditions after normalization was calculated using the corresponding bioconcentration factor value based on the National Food Safety Standard(GB 2762-2017 2017) for Cd in vegetables.The analyzed data are listed in Appendix K.

In the present study, a regression analysis was conducted between soil property parameters and HC5EDTA-addvalues to establish the predictive models of HC5EDTA-addas:

where HC5EDTA-addis the soil Cd concentration that would protect 95% of the cultivars (mg kg-1).

The derived calculations based on soil EDTAextractable Cd are presented as the continuous criteria in Table 2.Moreover, scenario thresholds calculated for different soil properties are also shown in Table 2.The Cd thresholds varied from 0.04 to 0.20 mg kg-1in soil with pH values ranging from 5.0 to 7.0.Because eq.(12) was based on the calculation of exogenous Cd, the native Cd of the soil should also be included.The concentration of EDTA-extractable Cd ranged from 0.01 mg kg-1to 0.21 mg kg-1in the 21 representative farmland soils in China(Appendix B).The arithmetic mean, geometric mean and median values of the EDTA-extractable Cd concentration were 0.07, 0.05 and 0.05 mg kg-1, respectively.In the present study, the median value (0.05 mg kg-1) was used as the default value.The comparison showed that the soil EDTA-extractable criteria were about 40 to 47% of the soil total Cd criteria.

Table 2 Soil EDTA-extractable Cd criteria for agricultural product safety

The predictive model for HC5EDTA-addwas applied to the field experiment data to test its validity and applicability (the data are summarized in Appendix L).The relationship between the predicted HC5EDTA-addvalues and the field experiment Cd concentrations is shown in Fig.6.Note that most of the predicted HC5EDTA-addvalues in the seven field experimental sites are below the 1:1 line, indicating the applicability of the predictive models.The predicted HC5EDTA-addvalues were mostly lower than the field experiment Cd concentrations, indicating that the predicted soil Cd concentrations could prevent the vegetables from exceeding the national food safety standard.Therefore, the established model was reliable for predicting the soil Cd concentration.

Fig.6 Comparison of predicted HC5EDTA-add values with the field soil EDTA-extractable Cd concentrations from different field sites.HC5EDTA-add, the soil EDTA-extractable Cd concentration for protecting 95% of the species.

4.Discussion

4.1.Soil criteria of total Cd for safe vegetable production

In this study, the soil thresholds of total Cd for safe vegetable production were calculated from predictive models which were derived from 90 matched soil and vegetable samples (Table 1).A generic BCF independent of soil metal concentrations and soil properties was used to build the predictive model for vegetables (Dinget al.2016).However, the generic BCFs were found to have the greatest uncertainty among the exposure factors in risk assessments.Since the bioaccumulation data were collected from the literature that included various soil properties, all the data should be normalized into the same soil properties in order to derive the criteria.The bioaccumulation data were normalized according to soil properties with the empirical soil-plant transfer model.After data normalization, the intra-species variabilityof BCFtotal-addvalues was reduced by 46.8 to 96.9%(Appendix E), which indicated that the empirical soil-plant transfer model was effective in reducing the uncertainty caused by the differences in soil properties.Therefore,the normalization removed the effects of soil properties to a certain extent.Subsequently, a soil property-specific BCF was calculated and applied in the predictive model.A similar approach has also been reported in previous studies (Liet al.2021).

In a previous study, the soil Cd thresholds for corrected aging time were found to be 1.5- to 2.2-fold higher than that with uncorrected aging time (Liet al.2021).Therefore, the aging effect was also taken into account in deriving the soil Cd thresholds for vegetables in this study.Since the data collected from the literature have different aging times, the data were normalized using an aging model.The normalization for aging time was effective in reducing the discrepancy caused by differences between laboratory and field conditions.One study claimed that the correction for aging time could avoid excessively conservative soil thresholds (Liet al.2018).

The soil Cd criteria for vegetables were derived using the SSD method approach.Wheeleret al.(2002)estimated that at least 10 to 15 data points are required for a reliable estimate of a particular endpoint.In this study, the data (90 cultivars of vegetables) met the above requirements.Therefore, the soil total Cd criteria for vegetables were reliable.

Currently, some studies on soil Cd thresholds for different kinds of vegetables have been conducted in China.According to the National Standard (GB 36783-2018), the soil Cd threshold values for rootstalk vegetables vary from 0.20 to 0.60 mg kg-1in soils with a pH of 5.0 to 7.0 and OC of 10 to 30 g kg-1.Our results were slightly higher than the results of the National Standard (GB 36783-2018) for rootstalk vegetables.The soil Cd thresholds for amaranth, celery, and pakchoi were calculated based on stepwise multiple regression models,with values of 0.26, 0.34, and 1.94 mg kg-1, respectively(Liet al.2019).According to the results reported by Luet al.(2017), the soil Cd thresholds for Chinese cabbage were in the range of 0.12-1.7 mg kg-1in different regions.Moreover, the soil Cd threshold for fruit vegetables based on the food pollution index was 1.2 mg kg-1in northern Shaanxi, China (Chenet al.2021).Since the absorption capacities of Cd differ among various vegetables, the soil Cd thresholds were different (Menget al.2021).If the soil Cd criteria for vegetables are variable, this is not beneficial for the management of vegetable production at a large-scale.Therefore, the soil total Cd criteria were established based on five types of vegetables, which covered almost all of the major vegetable types.

4.2.Soil EDTA-extractable criteria for vegetables

Most soil quality standards are still based on the total Cd concentration (Kim et al.2015).However, in the high geochemical background areas, such as karst areas,the total Cd concentration in the soil is high while the potential mobility and bioavailability of the Cd in soils is usually low (Wenet al.2020).For these reasons, the standard excess rate for the Cd concentration in edible parts of vegetables is lower than that in non-karst areas(Yanget al.2014; Wenet al.2020).Therefore, the soil EDTA-extractable Cd criteria were established for soil management practice.

Generally, the linear regression function is used to establish the predictive models for deriving the soil EDTA-extractable criteria for safer production (Kimet al.2020).However, due to the complex soil conditions at large spatial scales, the soil EDTA-extractable criteria are poorly correlated with Cd concentrations in the edible parts of vegetables (Yanget al.2020).Because of the robust applicability of the SSD method, it was applied to derive the soil EDTA-extractable criteria for vegetables in this study.The procedure for establishing the predictive model for soil available Cd criteria was the same as for the soil total Cd criteria.The EDTA-extractable Cd criteria for vegetables varied from 0.09 to 0.25 mg kg-1in soil with a pH of 5.0 to 7.0 and OC of 10 to 30 g kg-1(Table 2).

Legal standards of soil available Cd have been established in a few European countries (Kimet al.2015;Adamoet al.2017).However, the legal standard values vary among different countries.For example, in Australia and Germany, if the concentration of soil Cd extracted by 1 mol L-1NH4NO3in soil exceeds 0.04 mg kg-1, there is a potential risk of affecting the quality of food and forage crops.In Slovakia, the limit value of NH4NO3-extractable Cd in agricultural soil is 0.1 mg kg-1.In addition, the NH4NO3-extractable Cd limits are 0.4 and 1.0 mg kg-1for different soil textures in the Czech Republic.Generally,the soil available Cd concentration is estimated by various chemical extractions, such as 1 mol L-1NH4NO3, 0.01 mol L-1CaCl2, and 0.05 mol L-1EDTA (Kimet al.2015).But Hallet al.(1998) indicated that the use of 1 mol L-1NH4NO3extraction would underestimate the mobility of Cd in neutral to alkaline soils.The NH4NO3extractable Cd concentrations were often too low to be detectable(Kimet al.2020).Pueyoet al.(2004) showed that the Cd extraction rate obtained with the CaCl2procedure was generally lower than that obtained with the NH4NO3procedure.In addition, Menzieset al.(2007) has suggested that EDTA may provide a better evaluation of metal availability in heavily contaminated soil.Jianget al.(2020) also demonstrated that the relationships between EDTA-extractable and total added Cu provide a quantitative method for deriving the soil ecological criteria of available Cu.Therefore, the EDTA-extractable Cd concentration was used to indicate the EDTA-extractable Cd criteria in this study.Overall, the soil available Cd criteria may develop into a new solution for assessing the soil contamination risk for high geochemical background areas.

4.3.lnfluence of soil properties on the soil total Cd criteria and soil EDTA-extractable Cd criteria

Many studies have shown that soil pH and OC are the main factors affecting Cd bioavailability in soil (Yeet al.2014; Yanget al.2016).Soil pH has an influence on Cd speciation in the soil solution (Dinget al.2018).The increase in the soil OC concentration promotes the formation of stable organometallic complexes, thereby reducing the solubility of Cd in soil (Mohamedet al.2010).In this study, Fig.7 illustrates the growing tendency between the soil Cd threshold and soil properties.With increases in the soil pH and OC content, the values of the soil Cd threshold increased.However, the distance of the increase from they-axis for every pH unit increase was more than that for every OC unit increase.The parameterbof the Burr III distribution function based on soil properties also increased differentially for these two parameters (Appendix F).A partial correlation analysis was performed based on the calculated HC5total-addvalues for all 90 artificial soil scenarios, and the results showed that pH was the main variable affecting HC5addvalues,accounting for 95.6% of the variation in soil Cd HC5total-add,while OC explained only 29.5%.The variance tendency of the soil EDTA-extractable criteria based on soil properties was similar to that of the soil total Cd criteria, but the curve was relatively gradual.The parameterbof the Burr III distribution function for soil EDTA-extractable criteria was smaller than that for soil total Cd.Moreover, a partial correlation analysis based on the calculated HC5EDTA-addvalues was carried out, which showed that pH was the main variable affecting HC5EDTA-add(r=0.956), followed by soil OC (r=0.293).Therefore, both soil pH and OC should be considered when setting the soil total Cd criteria for vegetables.This view has also been confirmed in other research (Yanget al.2016; Dinget al.2018).

Fig.7 The HC5add values for soil total Cd (HC5total-add) and soil EDTA-extractable Cd (HC5EDTA-add) values under different soil conditions.

The soil EDTA-extractable Cd criteria were 40.0-46.7% of the soil total Cd criteria (Fig.7).Liet al.(2021)found that the extraction rate of soil extractable Cd was 39.6-77.4% when EDTA was used as the extraction agent, and the ratios of soil EDTA-extractable Cd criteria to soil total Cd criteria in this study were within that range.This view was also confirmed in research on the ecological thresholds for copper (Jianget al.2019).

5.Conclusion

The soil total Cd criteria were derived using the added risk approach based on data normalized with the soil properties and aging.The predictive model established in this study can accurately calculate continuous soil Cd criteria according to the soil properties.The soil total Cd criteria for vegetables were within 0.23 to 0.61 mg kg-1, and were affected by soil properties.Meanwhile,the predictive models for soil EDTA-extractable Cd were also established to reasonably assess the risks of soil Cd contamination in high background areas.The soil EDTA-extractable Cd criteria for vegetables were within 0.09 to 0.25 mg kg-1, and changed with soil properties.The results of this study indicated that the predicted and measured values mostly varied within the 95%confidence intervals.Overall, the criteria for soil total Cd and EDTA-extractable Cd derived in the present study will provide support for the risk assessment of soil Cd contamination for vegetable production in order to ensure food safety.

Acknowledgements

This study was funded by the Science and Technology Development Fund, Macau SAR, China (File 0159/2019/A3) and the National Key Research and Development Program of China (2016YFD0800406).

Declaration of competing interest

The authors declare that they have no conflict of interest.

Appendicesassociated with this paper are available on https://doi.org/10.1016/j.jia.2023.05.008