Prediction of Cadmium Transfer From Soil to Potato in Karst Soils, China

Contamination of food with the heavy metal Cd is a significant global concern. In this study, a field survey was performed to investigate the characteristics of Cd transfer from soil to potato tubers (n = 105). The results showed that the bioaccumulation factor of the potato tuber ranged from about 0.1 to 1. The soil threshold of Cd derived from the cumulative probability distribution was 0.15 mg kg−1 in order to protect 95% of potatoes. Additionally, prediction models for Cd transfer were constructed based on soil properties and the concentration of CaCl2-extractable soil Cd. The results of the analysis showed that pH was the critical factor affecting Cd uptake by potatoes. Additionally, the R 2 of different empirical models increased from 0.354 to 0.715 as the number of soil parameters was increased, and the predicted soil Cd concentration approached the measured values at values of about 0–15 mg kg−1. The results of this study suggest that the probability distribution method was stricter than the empirical prediction models for estimating the ecological risk of Cd contamination of potatoes in karst soils.


INTRODUCTION
With the development of industry and agriculture in China, contamination of food by the trace metal cadmium (Cd) has become a public concern due to its toxicity and persistence in soil. Soil Cd has negative effects on crop growth and production (e.g., wheat and rice) and can affect human health through the food chain (Diao, et al., 2005). Potatoes are an important food resource in China, which has the largest potato production in the world. In 2019, the Chinese potato planting area reached 4.7895 million hectares, and China's potato production of 91.938 million tons accounted for 24.91% of the global total. The consumption of potatoes that have accumulated Cd from contaminated soil is a major issue because potatoes are widely planted around the world. For example, excessive Cd accumulated in potatoes (about 0.18 mg kg −1 ) was shown to be caused by Cd pollution (1.25 mg kg −1 ) in sierozem soil in Gansu Province, China . Soil minerals in karst areas have high heavy metal content, and the metals in these minerals have been brought to the soil surface by geological and biological processes. Metal elements that have migrated to the surface have weak migration ability in the soil. Moreover, human activities in karst areas lead to a certain amount of heavy metal contamination in the local soil, further increasing the content of heavy metals and eventually leading to abnormal metal content (Tang, et al., 2021;Zhang, et al., 2021). Therefore, it is necessary to study the transfer of Cd from soil to crops in order to better control soil risks and ensure food safety.
In general, plant Cd concentrations are correlated with the soil total Cd. For example, soil total Cd accounted for 64% of the variation in the Cd concentration in potato tubers based on the results of stepwise multiple regression (n 49) when irrigation was performed with river water contaminated by mining (Carla, et al., 2007). However, McLaughlin et al., (1997) found no significant correlation between the Cd content in soil and that in potato tubers in a study of 352 potato production sites in Australia. These contradictory results may be due to differences in study regions, experimental conditions, and plant varieties. Soil total Cd concentrations can indicate the pollution status of soil, but cannot accurately reflect the pollution risk of soil to plants. The content of bioavailable Cd in soil mainly refers to the Cd that can be absorbed by crops during the crop growth period. In fact, many researchers have found that plant Cd content is closely related to the available Cd content in soil that is extractable by chemical extractants like CaCl 2 or CH 3 COONH 4 . The Tessier sequential extraction procedure can be used to extract five fractions of Cd with different reagents. For instance, Wei et al., 2020 used the fivestep batch metal fractionation method to study the influence of spent mushroom substrate on Cd immobilization and soil amendment, and Mitchell et al., 2020 employed a similar method to explore the impact of low-temperature biochar on the Cd concentration of calcareous river sediments. While single extraction methods seem to be more convenient and practical, Cd and Zn extracted by 0.01 M CaCl 2 were successfully used to predict Cd levels in brown rice planted in paddy fields across the western plains in Taiwan (Römkens, et al., 2009;Kara et al., 2004) assessed the soil Cd content in sowing regions of potato fields from a total of 45 different sampling stations in Turkey by using HCl + H 2 SO 4 . 0.01 M CaCl 2 is a commonly used chemical extractant in various extraction methods used to research the bioavailability of Cd because 0.01 M CaCl 2 solution matches the soil with respect to pH, concentration, and composition (Zhou, et al., 2019). Additionally, the physical and chemical properties of soil also have a great influence on the accumulation and absorption of Cd by crops. For example, soil pH and concentrations of Cl and Zn were found to be the main factors affecting the Cd concentration of potato tubers in Australia (McLaughlin et al., 1997). Furthermore, using a stepwise regression model, Rafiq et al., 2014 showed that the Cd phytoavailability of pak choi grown in Chinese soils was dependent on soil properties such as pH and organic matter content.
Empirical models such as Freundlich-type models have been widely used to study the accumulation and transfer of heavy metals between soil and plants. The Cd concentrations of plants and soil, as well as bioaccumulation factors (BAFs), could be added to Freundlich-type models in order to predict the phytoavailability of metals for the evaluation of environmental quality. For example, the Freundlich relationship was applied to the empirical modeling of soil parameters including clay content, soil organic matter, cation exchange capacity, and soil Cd to predict the Cd content in wheat grain (François, et al., 2009). Additionally, multiple linear regression models have been used to predict the metal content of plants. For instance, the uptake and transfer of Cd in potatoes (n 10) was predicted based on metal bioavailability (ethylenediamine tetraacetic acid extract) and soil properties in two contaminated regions of Kosovo (Zogaj and Düring, 2016). Moreover, Xu et al., 2019 used a polynomial model to derive Cd soil-plant relationships and soil criteria in order to provide suggestions for food safety and soil management. In recent years, the cumulative probability distribution method has been applied to study the soil threshold Cd concentrations in order to improve crop protection. For example, species sensitivity distribution (SSD) methodology is mainly used to determine the concentration of a toxicant that is protective of different ratios of species (50, 90, or 95%) in the environment. At present, the probability distribution functions used to fit the toxicological data of pollutants include log-normal, log-logistic, and Burr III type models. Liu et al., 2015 derived the soil threshold of Cd in a wheat-producing area of China (n 18) based on probability estimation using the log-normal function. Moreover, Ding et al., 2016 used the SSD to obtain the soil thresholds of Pb for root vegetables with a BAF-based prediction model depending on the combination of soil pH and cation exchange capacity (CEC). Additionally, Ding et al., 2018 applied a similar cumulative probability distribution (SSD) to determine the soil threshold of Cd based on 12 root vegetable cultivars fitted by a Burr type III distribution with three different Cd treatments in various types of soil. These studies show that linear regression models can be used to describe the transport and enrichment of heavy metals, and the relationships between heavy metals in soil and plants. Therefore, probability statistics such as the SSD method can be used to perform risk assessment of pollutants at the level of the whole ecosystem and set soil thresholds.
Paired soil and plant samples (n 105) were collected, and concentrations of Cd were measured. The present study aimed to 1) investigate the characteristics of Cd transfer from soil to potato tubers in a karst area of Guizhou Province, China, 2) obtain the soil Cd threshold based on the cumulative probability distribution using the log-normal function, and 3) explore an empirical prediction model based on plant and soil Cd concentrations and soil factors like pH and organic matter in order to improve soil environmental quality and protect food safety and human health.

Sample Collection
Field research was carried out in farmland in Guizhou Province, China. This province has a high background content of Cd due to the prevalent karst geochemical conditions, and the problem of soil Cd pollution is particularly prominent. For example, the relationships between soil properties and the accumulation of heavy metals in Brassica campestris L. in a karst area of Guizhou Province showed that the background levels of heavy metals in the soil of Guizhou Province are generally high due to the influence of topography and the presence of soil parent material ). An analysis of the risk posed by eight typical heavy metals in soils in Guizhou Province using province-wide data available in the literature revealed that Cd showed a very high potential ecological risk (Yu, et al., 2021). We selected the counties of Weining, Hezhang, Ceheng, Changshun, and Leigong, each of which contains a large proportion of carbonate rocks, for sampling. Soil was collected from the upper layer of the soil (0-20 cm). The distance between the sampling point and the local trunk road was more than 100 m.

Sample Analysis
The collected soil samples were air-dried and then crushed with wooden sticks before being passed through a nylon sieve with a diameter of 0.149 mm. The collected potato samples were washed with tap water to remove surface soil particles and were then washed with distilled water, put into an oven at 105-110°C for 15 min, and dried to a constant weight at 70-90°C in the oven. Finally, the samples were smashed for the subsequent analysis. The soil physical and chemical properties were determined by routine methods (Liu, et al., 2015), soil pH was determined by a glass electrode at a soil:water ratio of 1:2.5 (g ml −1 ), organic matter content was measured based on an oil bath heating method (potassium dichromate oxidation), and the results are shown in Table 1.
The total Cd in soil was determined as follows: we accurately weighed 0.1 g (accurate to 0.0001 g) of the sieved soil sample and placed it in a polytetrafluoroethylene (PTFE) tank, added 3 ml HNO 3 and 3 ml HCl, and placed the capped tank in a matched steel pipe, which was sealed. The steel pipe was placed in an incubator and heated at 185°C at a pressure of 790 kPa for 8 h. After cooling to room temperature, the digestion solution was removed from the pipe, placed on an electric heating plate, heated to 160°C, and steamed until it was viscous. The residual digestion solution was transferred to a 50 ml volumetric flask and diluted to 50 ml with ultrapure water.
Soil bioavailable Cd was determined by CaCl 2 extraction. A soil sample (2.00 g) was passed through a 0.2 mm sieve and placed in a 50 ml centrifuge tube. Then, 0.01 mol L −1 CaCl 2 solution was added and the tube was shaken for 2 h and centrifuged at 3,500 r·min −1 for 5 min. The level of Cd in the supernatant was detected. We accurately weighed 0.3 g (accurate to 0.0001 g) of each plant sample, placed it in a PTFE tank, and soaked it in 6 ml of HNO 3 overnight. Then, 2 ml of H 2 O 2 solution was mixed into the above solution and the tank was covered and placed in a supporting steel pipe and sealed. Plant Cd was determined as follows: the steel pipe was placed in an oven and heated at 100°C for 2 h. Next, the temperature was raised to 140°C for 2 h, after which it was raised to 160°C for 4 h. The method for solution transferal and detection of the plant Cd concentration was the same as that used for the soil samples. A total of 10 ml of each digested sample solution was analyzed by inductively coupled plasma mass spectrometry to determine its Cd concentration (ICP-MS; Thermo Fisher Scientific, Waltham, MA, United States, and x2). Reference material GSS-5 and GSB-1 was adopted to ensure the quality control and the detected Cd concentration in soil and plant deviated about 15% from the true value.

Modeling
Statistical analysis was performed using Microsoft Excel 2010 and SPSS 22.0 software. The log-normal function was applied to fit the cumulative probability distribution curve (SSD curve) using SSD Generator V1 software. The cumulative probability distribution was obtained from the toxicological data for pollutants based on the concept of BAFs (C soil 0.1/BAF, 0.1 mg kg −1 was the limit of the Chinese food safety standard (GB 2762(GB -2017 for Cd). The soil threshold values corresponding to different levels of protection or risk were obtained from the curve.
The BAF is the ratio of the metal concentration in plant tissues to that in the soil, and is calculated as follows: where C plant is the Cd concentration in the tuber of the test plant and C soil is the Cd concentration in the test soil. The relationships among soil Cd, plant Cd concentration, and soil properties were expressed by an empirical model as follows: Log C soil a × log C plant + b × pH + c × log other soil properties + k, where a, b, and c reflected the relationships among soil Cd, soil plant Cd, and soil properties, while k was the intercept.

Probability Statistics for Soil Cd Threshold
The total Cd concentration of the soil samples varied between 0.15 and 37.4 mg kg −1 , with an average value of 4.98 mg kg −1 , and the Cd concentration of the potato samples varied between 0.03 and 0.96 mg kg −1 , with an average value of 0.29 mg kg −1 ( Figure 1A). The soil and plant Cd concentrations generally exceeded the Chinese soil environmental quality risk control standard for soil contamination of agricultural land (GB 15618-2018; soil screening levels of 0.3 and 0.6 mg kg −1 , respectively, for pH ≤ 7.5 and pH > 7.5) and food hygiene standards (GB 2762-2017, <0.1 mg kg −1 ). The soil available Cd concentration varied between 0.0003 and 0.67 mg kg −1 , with an average value of 0.07 mg kg −1 ( Figure 1B). The extractable Cd was far less than the soil total Cd; the average soil total Cd concentration was about 71 times that of the average extractable Cd concentration. Although the total Cd concentration in the soil was high, only a small proportion of this Cd could be absorbed by plants. The variation in soil total Cd and soil available Cd concentration are shown in Figures 1A,B. Previous studies found that the soil Cd significantly affected the Cd content of plants; for example, Liu et al., 2010 observed a significant difference (p < 0.05) in the Cd concentration of 40 Chinese cabbage shoots (from 0.88 to 7.76 mg kg −1 ) when the soil Cd concentration increased from 1.0 to 5.0 mg kg −1 . However, Christopher et al., 2019 found that the Cd concentration in the flesh of potato was 0.2 mg kg −1 when the soil Cd concentration was lower than the detection limit. The results of experiments assessing the relationship between soil Cd content and plant Cd content have been inconsistent. Some studies have found that soil Cd levels were significantly related to plant Cd levels, while others have reported no relationship between these factors, perhaps due to the differences in plant species, climate, and other conditions. The variation in BAFs calculated by Eq. 1 was relatively large (about three orders of magnitude), facilitating the study of Cd transfer from soil to plants (Figure 2A). This large variation may be due to the fact that the external environment, including soil properties, plant varieties, irrigation, and fertilizer management, differed among the various study fields (Liu, et al., 2015). Statistical methods have been used to describe the probability distribution characteristics of BAFs in order to analyze Cd accumulation in potatoes. In the present study, it was found that the major BAF of Cd for plant uptake was in the range of 0.005-1.40, which is close to the BAF range found in a field survey of potatoes in four regions of New Zealand (Gray, et al., 2019). The critical value of the Cd concentration in soil may be deduced in terms of the limit of the Cd concentration in plants according to the measured bioaccumulation value (BAF Cplant/Csoil, Cplant is 0.1 mg kg −1 , which is the maximum value of Cd in potato tubers according to Chinese national food safety standards, and the corresponding Csoil is the soil Cd criteria) (Diao, et al., 2005). The soil Cd critical value of sampling soil sites can be used to fit the cumulative distribution curve of the soil threshold in order to prevent Cd contamination of potatoes in karst areas. In the present study, the soil Cd concentration that protected potatoes in 95% of soil sites was taken as the soil threshold of Cd. The resulting probability cumulative distribution can be built in many ways. For example, probability cumulative distribution functions, such as log-normal, log-logistic, or Burr type III, have been applied to study the cumulative distribution of the soil or water threshold of Cd. In general, it is necessary to estimate the potential risk of pollutants more conservatively for the risk assessment and derivation of soil environmental thresholds. In a previous study, the soil threshold of As for eight Chinese wheat varieties was calculated for 18 soils using cumulative probability distribution methods, including the Burr type III function, based on the protection of 95% of wheat varieties (Dai et al., 2016). In addition, hazardous concentrations of nonylphenol in the soil environment were estimated using probabilistic approaches based on acute and chronic species sensitivity distributions (Kwak, et al., 2017).
The cumulative probability distribution curve shown in Figure 2B characterizes the ecological risk to ecosystems; the potential affected fraction of Cd for the environment was calculated according to a given soil threshold Cd concentration (such as 0.3 or 0.6 mg kg −1 ). Furthermore, the distribution curve was also used to determine the soil threshold Cd concentration based on a given protection probability. For example, when the cumulative frequency was 20, 80% (100-20%) of soil sites have a safe soil Cd concentration below the soil Cd threshold. In the present study, it was found that the soil Cd threshold for the study area was 0.15 mg kg −1 to protect the plant samples in 95% of the soil sites based on the cumulative probability distribution (cumulative frequency 5%) ( Figure 2B). The cumulative probability distribution can also Frontiers in Environmental Science | www.frontiersin.org June 2021 | Volume 9 | Article 684887 be used to obtain other soil Cd thresholds for protecting different proportions of potatoes. The soil Cd threshold determined in this study as sufficient for protection of 95% of soil sites in karst areas was slightly lower than the Chinese soil environmental quality risk control standard for the contamination of agricultural soil (GB 15618-2018). These results indicate that potatoes grown in karst areas, in which tubers may more easily accumulate Cd from soil, may be more susceptible to soil Cd pollution in comparison with those grown in other areas.

Prediction Model for Soil and Plant Cd Concentration
Since the biochemical behavior of Cd in soil is complex, and many factors affect Cd uptake by plants, it is important to study the toxicity of Cd in mature crops in order to protect human health. Empirical regression can be used to estimate the rate of accumulation or transfer of heavy metals from soil to crops based on soil properties. For example, Liang et al., 2013 logtransformed the Cd content of spinach and soil to normalize variance and match a Freundlich-type function, and they added soil properties such as pH to the prediction model to improve its fitting accuracy. Liu et al., 2015 used a similar empirical method to fit a soil Cd threshold model using 18 Chinese soil types to prevent contamination of wheat by Cd. In the present study, we found that log Cd CaCl2 and log Cd total were both extremely significantly positively correlated with log Cd plant (p < 0.01). The Pearson correlation coefficient was 0.434 for both log Cd CaCl2 and log Cd plant , where it was 0.499 for log Cd total and log Cd plant ( Table 2). These results indicate that both the soil total Cd and the CaCl 2 -extracted Cd played an important role in Cd absorption by potato tubers. These findings are similar to those reported by Jun et al., 2018 from a study of the effects of alkaline amendment on Cd bioavailability in two Chinese cultivars of polished rice (Xiangwanxian 13 and Zhongyou 9,918) grown in paddy soil. A recent study using single extraction procedures, including methods utilizing EDTA, DTPA, CH 3 COOH, and HCl, revealed that the extractable Cd in soil was linearly correlated with the Cd concentration in wheat grains . These results are consistent with our findings, which showed a similar relationship between Cd in soil and plant samples. A stepwise linear regression was performed to establish the relationship between log Cd soil (soil total Cd), log Cd CaCl2 (soil CaCl 2 -extracted Cd), and soil pH and organic matter (OM) (n 105) based on Eq. 2. The results of the prediction models showed that Cd total was significantly correlated with the soil pH and OM, soil-extractable Cd CaCl2 , and Cd plant ; the R 2 value of the prediction model increased from 0.354 to 0.715 as the number of independent variables was increased (Table 3). Linear regression equations were obtained to describe the relationships between the total Cd concentration in soil and the Cd concentrations in wheat grains and roots (R 2 0.62, 0.71 for grains and roots, respectively), as well as between the CaCl 2 -extracted Cd concentration in soil and the Cd concentrations in wheat grains and roots (R 2 0.58, 0.49 for grains and roots, respectively). A recent study indicated that the soil total Cd and CaCl 2 -extracted Cd were useful predictors of the Cd transfer from soil to plants (Qu, et al., 2020), which was in agreement with the results described in the present study. Soil pH was an important predictive factor, as shown by its presence in every prediction model in Table 3, and, among the independent variables in equation (4), pH contributed approximately 47% of the variance in the dependent variables.
Prediction of soil threshold Cd values by linear equations is more convenient than prediction by the cumulative probability analysis method. The soil threshold calculated by the empirical prediction equation was greater than that calculated by the probability statistic method. For example, the soil thresholds of Cd were 0.35, 0.97, and 1.60 mg kg −1 when the soil pH was 5.5, 6.5, and 7, respectively, and the soil CaCl2 , OM, and Cd plant were 0.1 mg kg −1 , 10 g kg −1 , and 0.1 mg kg −1 , according to equation (4) in Table 3. Previous studies have shown that soil properties, especially pH, are crucial factors influencing the transport of Cd between soil and vegetables. pH mainly affects the equilibrium distribution of heavy metal ions at the two-phase interface of soil and water, as well as the formation and dissolution of soil carbonate. Soil organic matter has a large number of functional groups, and its ability to adsorb Cd is much higher than that of other mineral colloids. Humic acid formed by decomposition can chelate (complex) with heavy metals, inhibiting the absorption of metals by plants. Using a stepwise regression model, Rafiq et al., 2014 concluded that Cd phytoavailability to pak choi was significantly affected by soil pH, organic matter, and total Zn and Cd concentrations, and the R 2 of the stepwise regression model in their report was 0.977. Furthermore, through stepwise forward multiple regressions, McLaughlin et al. (1997) found that soil parameters such as pH and Cl were important influences on the Cd concentration of potato tubers. Furthermore, the measured soil Cd concentrations in the present study were close to the values predicted using equation (4) ( Table 3) for a certain soil Cd concentration range (about 0-15 mg kg −1 ). The predicted soil Cd concentrations were greater than the measured values for higher soil Cd concentrations (Figure 3). In contrast with our results, a recent study reported no significant relationships between soil properties and the Cd concentration in potatoes in three field sites across New Zealand (Gray, et al., 2019). A recent study in Hunan province reported a soil Cd threshold of 1.819-3.272 mg kg −1 based on the safety limit of Cd in cereals (0.2 mg kg −1 ), which was higher than that permitted by the national soil environmental standards of China (Chen, et al., 2019). Using a prediction model, the threshold of soil total Cd in rootstalk vegetable fields in Guangdong was calculated to be 0.94 mg kg −1 , which was also higher than the national soil Cd threshold in China (Sun, et al., 2013). Although the threshold values of Cd from studies performed in non-karst area were higher than those from our research, these differences are likely due to differences in soil conditions, pollution conditions, plant species, and other factors. Further research is required to compare Cd uptake rates and soil threshold values for karst and nonkarst areas.

CONCLUSION
In the frequency distribution of the BAF of potato tubers, BAF values from 0.1 to 0.3 occurred relatively frequently for soil Cd concentrations ranging from 0.15 to 37.4 mg kg −1 . The transfer of Cd from soil with a high Cd concentration to potato tubers was found to be highly positively correlated with soil pH, OM, and CaCl 2 -extractable Cd (R 2 0.715). The log-transformed plant Cd concentration was highly significantly correlated with the soil total Cd and soil CaCl 2 -extractable Cd concentrations, with Pearson correlations of 0.499 and 0.434, respectively. The log-normal function was used to establish the cumulative probability distribution of the soil Cd threshold. Our results show that the probability model and empirical model can each adequately describe the transfer of Cd from soil to crops.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, and further inquiries can be directed to the corresponding author.

AUTHOR CONTRIBUTIONS
KL analyzed the data and wrote the manuscript, HL revised the manuscript, XZ collected the samples, and ZC and XW provided important suggestions for the manuscript.