Improved prediction of the bioconcentration factors of organic contaminants from soils into plant/crop roots by related physicochemical parameters.

There has been an on-going pursuit for relations between the levels of chemicals in plants/crops and the source levels in soil or water in order to address impacts of toxic substances on human health and ecological quality. In this research, we applied the quasi-equilibrium partition model to analyze the relations for nonionic organic contaminants between plant/crop roots and external soil/water media. The model relates the in-situ root concentration factors of chemicals from external water into plant/crop roots (RCF(water)) with the system physicochemical parameters and the chemical quasi-equilibrium states with plant/crop roots (αpt, ≤1). With known RCF(water) values, root lipid contents (flip), and octanol-water Kow's, the chemical-plant αpt values and their ranges of variation at given flipKow could be calculated. Because of the inherent relation between αpt and flipKow, a highly distinct correlation emerges between log RCF(water) and log flipKow (R2 = 0.825; n = 368), with the supporting data drawn from 19 disparate soil-plant studies covering some 6 orders of magnitude in flipKow and 4 orders of magnitude in RCF(water). This correlation performs far better than any relationship previously developed for predicting the contamination levels of pesticides and toxic organic chemicals in plant/crop roots for assessing risks on food safety.

Introduction

Since the advent of pesticides in 1950s for the control of insects and weeds, there has been a strong demand for a basic understanding of the transfer of pesticides and toxic chemicals from soils to food crops (Harris and Sans, 1967; Iwata et al., 1974; King et al., 1965; Fuhremann and Lichtenstein, 1980). Such knowledge is needed for estimating the levels of harmful substances in food chains and the associated human health impacts. To meet this demand, extensive greenhouse and field-plot studies were performed by scientists (Beestman et al., 1969; Harris and Sans, 1967; Lichtenstein, 1959) to explore relations between the levels of pesticides in soils (Cs) and the levels absorbed by crops (Cpt). The crop samples examined included whole plants and their roots and shoots. These studies revealed that: when a pesticide in a soil was absorbed by a crop species or its segment (e.g., root) at a fixed system setting, a relatively good linear relation existed between Cpt and Cs; however, the Cpt/Cs ratio of a chemical with a crop varied widely between soils. In spite of these findings, the soil-to-plant bioconcentration factors, i.e., BCF(soil) = Cpt/Cs, are routinely measured as reference indices.

Investigation on relations between levels of organic chemicals in plants/crops and in external sources (soil or water) receives increasing attention since 1970s as more varieties and higher quantities of chemicals are being emitted into natural systems that arouses concerns for food safety. In the 1990s, theoretical models were being developed by Riederer (1990), Trapp et al. (1990), Paterson et al. (1994), Trapp and Matthies (1995), and Burken and Schnoor (1998) for plant uptakes of chemicals in terms of their interfacial transfer rates and partition constants between plants and associated media: soil, water, and air. The models were executed with estimated transfer rates and coefficients along with estimated plant transpiration and chemical-breakdown rates. Later, Chiou et al. (2001) formulated a quasi-equilibrium partition model by relating the levels of chemicals in a plant to their quasi-equilibrium partitions to the associated plant components, with the quasi-equilibrium values (αpt ≤ 1) depending on chemical properties and system settings. The latter model was substantiated by the disparate quasi-equilibrium states of different chemicals in a plant (or its segment) following an exposure. The model has since been further tested and verified by many water-plant and soil-plant contamination data (Barbour et al., 2005; Card et al., 2012; Chiou, 2002; Gao et al., 2005; Huang et al., 2011; Li et al., 2005; Li et al., 2002; Su and Zhu, 2006; Su et al., 2009; Yang et al., 2017; Zhang et al., 2005).

A critical parameter to be dealt in the quasi-equilibrium partition model is the quasi-equilibrium αpt value, which is a function of the chemical’s property (Kow) and the plant-uptake setting. Since the ranges of variation in αpt for chemicals under diverse system settings have not been established (Chiou et al., 2001), we here extended the basic model to analyze the ranges of αpt values for various chemicals with plant/ crop roots based on the observed root concentration factors (RCFs) from extensive laboratory/greenhouse/field studies. Incorporating the observed αpt ranges into the model enables one to construct helpful correlations of RCFs with relevant parameters to aid in estimation of the crop-contamination level that is timely needed. This objective is further motivated by findings that when the driving force of a chemical from soil to plant roots is based on the level in soil pore water (Cpw), instead of the level in bulk soil (Cs), the chemical RCFs with roots from extensive sources converge sharply. A brief overview of the quasi-equilibrium partition model is given next to facilitate the subsequent account of extensive literature data toward the formation of improved predictive correlations.

Theoretical considerations

The fact that no general relation exists between the levels of chemicals in plants (Cpt) and sustaining soils (Cs) suggests that Cs is not a rigorous intensity index (Hung et al., 2009). It is known that the sorption of relatively water-insoluble nonionic organic chemicals by a water-saturated soil occurs predominantly by partition into the soil organic matter (SOM) (Chiou et al., 1979); adsorption by soil minerals is relatively insignificant except for soils having very low SOM contents because of the adsorptive competition by water (Chiou et al., 1981, 1983). Therefore, the level of a chemical in soil pore water (i.e., Cpw), which serves as the driving force for plant uptake, depends primarily on the level in SOM (i.e., Com), not the level in bulk soil (Cs), i.e., where Kom (= Com / Cpw) is the chemical’s equilibrium partition coefficient between SOM and water and fom is the SOM fraction in soil. Here Kom is related to the soil-water distribution coefficient (Kd) by

The term Com in Eq. (1) expresses the contamination intensity in a water-saturated soil as sensed by the equilibrium Cpw in soil pore water. Since Cs (= fomCom) varies not only with Cpw but also with fom, the term RCF(soil) based on Cs is elusive. The finding (Kile et al., 1995) that the Kom of a chemical is practically constant between soils from wide sources enables the Cpw with various soils to be rapidly estimated via Eq. (1).

To analyze the soil-plant contamination data, the needed Cpw values could thus be derived from Cs, fom, and estimated Kom using the above procedure, if the soil-specific experimental Kd values are not available. Using the Cpw as the driving force in soils, the levels of chemicals absorbed by a plant segment after an exposure could be set as: where Cpt is the level in a plant segment on the fresh-weight basis; fpw is the weight fraction of water in the plant segment; fpom is the weight fraction of all organic constituents in the plant segment (with fpw + fpom = 1); Kpom is the partition coefficient of a chemical between the (dry) plant organic matter and water; and αpt is the quasi-equilibrium factor expressing the deviation from equilibrium of a chemical in a plant segment relative to its level in external water when the plant is sampled. The αpt value (≤ 1) has a logical upper limit of 1 when the chemical in a plant segment of interest reaches equilibrium with its external water phase.

For simplicity, the most important plant constituents are assumed to be water, carbohydrates, and lipids, such that Eq. (3) may be expressed as where subscripts “ch” and “lip” refer respectively to carbohydrates and lipids, with fpw + fch + flip = 1. Since experimental Cpw data are usually unavailable, they are estimated from Com/Kom via Eq. (1) to give

Thus, the in-situ root concentration factor of a chemical relative to its level in external soil water (Cpw), RCF(water), may thus be defined as: or alternatively as

The RCF(water) as defined is a quasi-equilibrium distribution coefficient. Unlike many equilibrium BCFs, the RCF(water) depends not only on the chemical-plant properties but also on the αpt (≤1) when soils and plants are collected.

Under a system setting, the αpt may be calculated via either Eq. (6) based on the measured fpw, fpom, and Kpom or via Eq. (7) using available fpw, fchKch, and flipKlip. Generally, if fpomKpom or flipKlip is only a small multiple of fpw (~0.9), as for chemicals with Kow < 500 and roots with flip < 0.01, the αpt should be close to 1 after short plant exposures (Briggs et al., 1982; Chiou et al., 2001). Here the fchKch term is usually small relative to flipKlip (Hung et al., 2010) such that fchKch could be neglected in most cases. If flipKlip/fpw is large (say, > 50–100), the αpt value will be small and highly system-dependent, as described later. In the latter case, it is practical to determine the ranges of αpt for common pesticides and contaminants on food crops under assumed system settings. Such information serves to estimate the BCF(water) ranges and contamination levels of the chemicals in crops.

Data processing and manipulation

To analyze the RCF(water) of chemicals in Eq. (7), the needed Kom data consist of either the experimental values or those estimated from Kow using the empirical equations described below. For relatively nonpolar substituted aromatic compounds other than polycyclic aromatic hydrocarbons (PAHs), the log Kom-log Kow correlation of Chiou et al. (1983) is used to estimate the soil Kom:

For PAHs and their derivatives, the original log Koc-log Kow correlation established by Karickhoff et al. (1979) with sediments is converted to that for soils using the relation of Koc (sediment) ≅ 2 Koc (soil) (Kile et al., 1995) along with Koc = 1.85Kom (Chiou, 2002) to give:

For polar chemicals, the empirical correlation of Briggs (1981) could be used for estimation if no experimental Kom are available:

When the reported RCF(soil) or RCF(water) data of chemicals are based on the dry plant weight, they are converted to the fresh-weight values for consistency of the treatment. This conversion is made by assuming the water content to be 90% by weight for all fresh roots. This water content is the average for the edible portions of 172 crops/plants given in the USDA Nutrient Database (https://ndb.nal.usda.gov/ndb/search/list).

As indicated, the fchKch term in Eqs. (5) and (7) is generally small relative to flipKlip. The value of Kch may be estimated from the respective Kow via the correlation (Hung et al., 2010):

Finally, the Klip term in Eqs. (5) and (7) is assumed to be equal to Kow for the model execution because of the lack of Klip data for most chemicals. This assumption is justified by the similarity of Klip and Kow for a diversity of chemicals according to Hung et al. (2014):

As seen, the difference between log Klip and log Kow is not a serious concern to the predicted log RCF(water) for all chemicals.

Sources of literature data

The RCF(water) data of 37 chemicals with a number of plants/crops from 11 hydroponic studies (Briggs et al., 1982; De Carvalho et al., 2007; Gao et al., 2008; García-Valcárcel et al., 2016; Hinman and Klaine, 1992; Namiki et al., 2015; Romeh, 2014; San Miguel et al., 2013; Su and Zhu, 2006; Su et al., 2009; Xia and Ma, 2006) are listed in Table S1 (Supporting Information). Total data points are 48. Plant roots studied include rice seedlings, lettuce, parrot feather, red clover, Hydrilla verticillata, Phragmites australis, Hordeum vulgare, Brassica aleracea, barley, and many others. Chemicals studied include industrial wastes (e.g., benzaldehyde, chlorophenols, and chlorobenenes), various pesticides (e.g., atrazine, oxamyl, aldicarb, phenylureas, carbamates, chlordane, ethion, lindane, dieldrin, etc.), and pharmaceuticals (e.g., fluconazole and clotrimazole), with the log Kow = −0.57 to 5.41.

A similar compilation of the RCF(water) data for 66 chemicals from various soils to various plants/crops is listed in Table S2. Total data points are 376. The compiled values are from 4 field experiments (Harris and Sans, 1967; Kipopoulou et al., 1999; Mikes et al., 2009; Zhang et al., 2005) and 15 laboratory/greenhouse studies (Beestman et al., 1969; Boxall et al., 2006; Cai et al., 2008; Carter et al., 2014; Gao et al., 2005; Huang et al., 2011; Jiang et al., 2016; Macherius et al., 2012; Pannu et al., 2012; Prosser et al., 2014; Scheunert et al., 1994; Tao et al., 2009; Trapp et al., 1990; Wu et al., 2012; Zhu et al., 2016). The source data comprise a wide range of log Kow (0.91–8.70), plants/ crops, and soil fom values. Major studied chemicals include organochlorine pesticides (Beestman et al., 1969; Harris and Sans, 1967; Mikes et al., 2009), chlorinated benzenes (Scheunert et al., 1994; Trapp et al., 1990; Zhang et al., 2005), PAHs (Cai et al., 2008; Gao et al., 2005; Kipopoulou et al., 1999; Tao et al., 2009), and poly-brominated diphenyl ethers (PBDEs) (Huang et al., 2011), which make for > 90% of the total data points. Minor chemical classes include polychlorinated biphenyls (PCBs) (Mikes et al., 2009) and personal-care products (Macherius et al., 2012; Pannu et al., 2012; Prosser et al., 2014; Wu et al., 2012). Studied plant roots and their lipid contents (flipx100) are wheat (1.14), barley (1.00), carrot (0.24), radish (0.10), celery (0.17), maize (0.53), pumpkin (0.70), turnip (0.10), onion (0.10), spinach (0.34), Chinese cabbage (0.68), ryegrass (0.32), and amaranth (0.32). If the flip data are not reported in original sources, available values from the USDA Food Composition Database (https://ndb.nal.usda.gov/ndb/search/list ) are selected. Also listed in Table S2 are the log Kow, soil organic-matter contents (fom), plant species, root lipid contents (flip), and plant exposure times. The sources of log Kow values and experimental or model-estimated log Kom values of chemicals in Table S2 are indicated in Table S3.

The hydroponic RCF(water) data selected for analysis are restricted to the plant samples with > 1-day exposure. For the soil-plant RCF(water) data from laboratory or field studies, only the data for samples with > 7 days (up to 93 days) of exposure were selected, with most data (~85%) having exposure times of 39–70 days. In either case, no RCF(water) data of ionic and ionizable chemicals were selected since their uptakes by plants may involve more than a partition process (Tanoue et al., 2012; Wu et al., 2013; Li et al., 2019). When adopting the soil-plant data from literature, only the studies with reported soil fom and accessible plant flip values were selected. For given chemicals with different soils or plants, or with same soils and plants but at very different dosed levels in soil or very different exposure times, the associated RCF(water) values were treated as independent data in the correlation analysis.

Results and discussion

Root concentration factors from water solutions

We evaluate first the RCF(water) values of chemicals, i.e., the levels in fresh roots (Cpt) relative to those in water (Cw), from 11 hydroponic studies listed in Table S1. As shown in Fig. 1, a plot of log RCF(water) versus log Kow of the chemicals creates two distinctive groups. Over the range of log Kow ≥ 1.5, a good linear relation is observed between log RCF(water) and log Kow, i.e., with a correlation coefficient R2 = 0.908. The 95% confidence interval and prediction interval for the RCF(water) data are drawn in Fig. 1. Conversely, in the range of log Kow ≤ 1.1, the data fall close to the horizontal line of RCF(water) ≅ 0.9, as first noted by Briggs et al. (1982). The latter results are much anticipated because when the Kow of a chemical is small the flipKlip will be small compared to fpw (which is ~0.9) if flip is small. In this case, log RCF(water) approaches log fpw, or essentially zero, as the lower limit. As shown in Table S2, the flip values for most fresh roots are < 0.01 (i.e., 1%) such that for all chemicals with Kow ≤ 10 the log RCF(water) is practically zero.

Fig. 1.

Plots of log RCF(w) versus log K for organic chemicals on plant/crop roots from 11 hydroponic studies described in the text. The “expected line” to the lower left is based on Eq. (7) with αpt = 1, fpw = 0.9, fchKch = 0, and flipKow = 0.

The log RCF(water) at a log Kow ≥ 1.5 displays a moderate scatter due to the different root flip and αpt values as impacted by the chemical Klip, plant type, plant growth/size, exposure time, and system setting. Whereas calculations of the αpt values for individual chemicals with roots in Table S1 require the flip data that are inaccessible, except the flip (~0.01) for barley roots (Briggs et al., 1982), one expects the αpt value to decrease largely with increasing Kow, as found with limited data earlier (Chiou et al., 2001). In principle, with Kow < 500 or flipKlip < 5, the αpt should be generally close to 1. If the flip is very small (e.g., ~0.1% for radishes), even the uptakes of chemicals with Kow = 1000–2000 could approach their equilibrium limits (i.e., αpt = 1) in short times. In contrast, the RCF(water) values of high-Kow chemicals tend to be much more system-dependent. In later analyses of diverse soil-plant data with known flip values, the observed pattern between αpt and flipKow reflects the impacts of various system parameters on RCF(water).

RCFs based on chemical levels in soils and SOM

Consider next the RCF(soil) values based on the levels of chemicals in plant roots on a fresh-weight basis (Cpt) and the corresponding levels in whole soils on a dry-weight basis (Cs). The relevant data are listed in Table S2, the SI. As noticed, when log RCF(soil) is plotted against log Kow in Fig. 2(A), no distinct patterns emerge (R2 < 0.16), as recognized in various reports (Cai et al., 2008; Doucette et al., 2018; Macherius et al., 2012; Mikes et al., 2009; Scheunert et al., 1994; Takaki et al., 2014; Tao et al., 2009; Trapp et al., 1990; Wu et al., 2012). Here the RCF(soil) at a given Kow varies by as high as 1400 -fold. In addition, many observed RCF(soil) values are < 1, which is in contrast with the trend for organic chemicals concentrating from water into an organic phase. In analyzing a large set of chemical soil-plant data, Doucette et al. (2018) and Takaki et al. (2014) found a similar wide scatter in plots of log RCF(soil) versus log Kow. The lack of a specific relation between log RCF(soil) and log Kow comes with no surprise in view of Cs = fomCom for relatively water-insoluble (high-Kow) chemicals with a water-saturated soil, where the soil contamination intensity Com is independent of the value of fom (see more discussion in Hung et al., 2009). Since the fom varies between soils, the influence of fom alone on RCF(soil) could well exceed two orders of magnitude in extreme cases. Thus, the observed RCF(soil) of a high-Kow chemical with a plant tends to be inversely related to soil fom as manifested by the dieldrin RCF(soil) data (Beestman et al., 1969; Harris and Sans, 1967). Moreover, the RCF(soil) of a chemical with a given soil tends to increase with increasing plant flip, although not in a strictly proportional manner.

Fig. 2.

Plots of log RCF(soil) versus log Kow (A) and log RCF(water) versus log Kow (B) for organic chemicals on plant/crop roots from 19 soil-plant studies described in the text.

With above considerations, it is expected that if the distribution of chemicals between soil and plant is expressed in terms of the levels in SOM (i.e., Com), the resulting log RCF(som) values should correlate better with the respective log Kow. The needed RCF(som) data are obtained by multiplying the RCF(soil) data by associated fom values (see Eqs. (1) and (7)). The plot of log RCF(som) versus log Kow is shown in Fig. S1, the SI. As seen, although this adjustment does significantly enhance the correlation (R2 = 0.421) with a reduced RCF(som) spread, the correlation coefficient is not high enough to be practically useful. In essence, the log RCF(som)-log Kow plot exhibits a relatively small negative slope (−0.282), implicating that other factors and system parameters also contribute significantly to in-situ RCF values.

RCFs based on chemical levels in soil water

Recognizing that the RCF(soil) is not a rigorous index and the SOMbased RCF(som) offers only a moderate improvement, we now examine the ability of the soil-water-based RCF(water) in Eq. (7) to reconcile the same set of data in Table S2. The desired RCF(water) data can be obtained by multiplying the RCF(soil) by respective fomKom. As shown in Fig. 2(B), a plot of log RCF(water) versus log Kow for the soil-plant data generates a reasonably distinct and positive correlation, i.e., with a correlation coefficient R2 = 0.700, which reduces enormously the RCF(soil) scatter in Fig. 2(A). This improvement signifies the merit of RCF(water) values based on chemical levels in soil water, which serves to link the RCF(water) data derived from soils and water solutions. The observed spread in RCF(water) at a given Kow reflects the combined impacts of the root flip and chemical-root αpt, which are a function of plant species, plant growth, plant health, exposure time, and system variables, when the soil-plant samples are collected.

Based on the model, the plant flip is a prime factor to the RCF(water) of a chemical whenever the flipKow far exceeds fpw (~0.9). This expectation is substantiated by a further vastly improved correlation of log RCF(water) with associated log flipKow, as shown in Fig. 3, where the data split into two groups. For the group with log flipKow < 0, the observed log RCF(water) approaching 0 at the lower end is well in line with Eq. (7) for many low-Kow chemicals, as shown in Fig. 1. Conversely, for the group at log flipKow > 0, the log RCF(water) data, except those of hexachlorocyclohexanes (HCHs) discussed later, are eminently linearly related to the log flipKow to give with a correlation coefficient R2 = 0.825, which covers about 6 orders of magnitude in flipKow and 4 orders of magnitude in RCF(water). The associated 95% confidence interval and prediction interval are also shown. The success of Eq. (15) is due to a large extent to the inverse relation between αpt and flipKow to be illustrated later. As is expected, the log RCF(water) values (−0.18 to 1.91) of 18 chemicals (with log Kow = −0.57 to 4.6) from water into barley (flip = ~0.01) (Briggs et al., 1982) in Fig. 1 fit closely the plot in Fig. 3.

Fig. 3.

Plots of log RCF(water) versus log [flipKow] for organic chemicals on plant/crop roots from 19 soil-plant studies described in the text. The solid-redcircle data points are suspected outliers excluded from the regression analysis. The “expected line” to the lower left is based on Eq. (7) with αpt = 1, fpw = 0.9, and fchKch = 0. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Besides the sharply improved fit, the plot in Fig. 3 helps to identify the anomalous data, i.e., those which fall outside the model-acceptable range. First, if the reported RCF(water) values are far < 1, or the log RCF(water) far below zero, following a long plant/crop exposure, they must be viewed as suspect, because the theoretical lower limit of normal log RCF(water) according to Eq. (7) is where RCF(water) = fpw = ~0.9 with αpt = 1. Similarly, if the RCF(water) data give αpt ≫ 1, such data are usually not justified. For chemicals in roots spanning a large range of flipKow/fpw, the theory predicts the αpt to decrease with increasing flipKow. Calculated αpt values of chemicals with studied roots via Eq. (7) (using fpw = 0.9) are listed in Table S2. Using the reported flip data, the spread in log RCF(water) at a given log flipKow in Fig. 3 reflects the degrees of deviation of the αpt values of all chemical-root pairs having same flipKow from their equilibrium limits (αpt = 1) under the assumed system settings. A plot of chemical-root αpt values against respective flipKow values is displayed in Fig. 4.

Fig. 4.

Plots of αpt versus log [flipKow] for organic chemicals on plant/crop roots from 19 soil-plant studies described in the text (n = 376).

Dependence of αpt on chemical lipophilicity and system setting

In Fig. 4, the mean αpt value at a flipKow shows an expected inverse relation with flipKow, decreasing from about 1 for many chemicals at log flipKow ≤ 0, or flipKow ≤ 1, to about 0.001–0.002 at log flipKow = 6.5 or flipKow = 3 × 106 for highly brominated PBDEs. The observed positive slope between log RCF(water) and log flipKow at log flipKow > 0 in Fig. 3 implies that the decrease in mean αpt with increasing flipKow is less than the net increase in flipKow (or flipKow). As manifested in Fig. 4, the log αpt-log flipKow relation over log flipKow = −2.0 to ~6.0 is clearly nonlinear, consistent with the quasi-equilibrium partition model. Only at log flipKow ≥ 1, the log αpt-log flipKow plot is approximately linear, similar to the largely linear relation of log RCF(water) versus log flipKow at log flipKow > 0. Thus, a simple algorithm to find the (approximate) mean αpt at any log flipKow ≥ 1 is to first find the mean RCF(water) via Eq. (15) and then substitute it into Eq. (7) to calculate the mean αpt. Calculations of αpt may be expedited with fpw = 0.9 and fchKch = 0 without causing significant errors. Here the calculated mean αpt values for chemicals in roots (Table S2) at log flipKow = 1, 2, 3, and 4, for example, are 0.53, 0.22, 0.085, and 0.033, respectively, under the assumed exposure conditions.

In theory, the width or spread in αpt and RCF(water) at a flipKow should diminish with reducing flipKow because the related mean αpt would be closer to the upper limit of 1; therefore, the log RCF(water)-log flipKow or logαpt-log flipKow relation generally becomes sharper at reduced flipKow. On this basis, the unusually small RCF(water) (0.15 to 0.35) or αpt (0.024–0.056) values of α-, β-, and γ-HCH on radishes, which result from an unusually large spread in RCF(water) at small flipKow (5.2–6.5) in a field study (Mikes et al., 2009), are considered abnormal as well as the RCF(water) of 1,4-dichlorobenzene on barley (Scheunert et al., 1994), which yields αpt = ~2. Excluding these outliers, the αpt or RCF(water) values at given flipKow differ only by a factor of < 2 to 30, which is rather small considering that the RCF(water) values encompass a wide variety of chemicals under a wide diversity of system settings.

It is beneficial to pinpoint essential factors affecting the αpt value. Among all potential factors, accuracies of Kow and Kom data have a direct impact on αpt especially for high-Kow chemicals (e.g., PAHs, PCBs, and PBDEs). In addition, calculations of Cpw via Com/Kom tend to overestimate Cpw for chemicals with high flipKow/fpw and thus underestimate the actual RCF(water) and αpt. This is because the model assumes Cpw to be in instant equilibrium with Com during the plant uptake, which is doubtfully valid when flipKow is large. There is, however, a benefit gained by using the estimated Kom data for homologous chemicals because the data usually yield right relative orders to enhance the relative consistency of the estimated RCF(water). In theory, the extent of underestimation in αpt increases with increasing flipKow.

Based on the model, the plant flip is a major factor on the RCF(water) of lipophilic chemicals. When assessing the flip impact on RCF(water), the root flip is assumed, as approximation, to be independent of the plant size, plant growth/health, and system settings. This assumption introduces certain uncertainties in RCF(water). Other factors, e.g., the magnitude and uniformity of fom and Com and the plant/crop exposure time, could also affect the observed RCF(water) or αpt values. Accounts for some of these factors based on relevant available soil-plant contamination data are presented next.

Influences of soil chemical level and plant exposure time

The potential impacts of Cs and the plant exposure time on RCF(water) are of immediate interest. When the studied soils are well homogenized in a laboratory study, the resulting RCF(water) and αpt values of dieldrin with maize roots (Beestman et al., 1969) show only a small dependence (by a factor of < 2) on Cs (from 1 and 5 mg/kg) and on the time of exposure (39 to 90 days). Results of Harris and Sans (1967) on dieldrin uptakes by carrots from three well-mixed soils in well-maintained field plots over a period of 93 days also yield comparable RCF(water) and αpt values despite that the soil fom varies by nearly 50-fold (0.014 to 0.665). These findings disclose that the chemical-plant αpt values are not sensitively affected by Cs and fom or the exposure period if the time of exposure is sufficiently long and if the soil fom (or chemical Com) is kept relatively uniformly.

Influences of plant species and system settings

In the preceding reports of Beestman et al. (1969) and Harris and Sans (1967), the observed αpt values of dieldrin (i.e., 0.4–0.6 and 0.2, respectively) are however notably different. It suggests that system settings and plant species significantly affect the αpt values of chemicals with large flipKlip/fpw in plants. This view is supported by extensive data on the root uptakes of phenanthrene (PHN) and other PAHs under different (laboratory and field) settings over exposure times of 45–64 days (Cai et al., 2008; Gao et al., 2005; Kipopoulou et al., 1999; Tao et al., 2009). Here, the αpt or RCF(water) values of PHN and pyrene taken up from a soil at 7–8 different initial PHN or pyrene levels (Cs or Com) by a plant (e.g., Chinese cabbage) in a laboratory differ only by about 2-fold (Gao et al., 2005). In contrast, the RCF(water) values of PHN and pyrene with different plants and system settings differ by 30 to 40fold (Gao et al., 2005; Kipopoulou et al., 1999; Tao et al., 2009), in which the flip varies by about 5-fold and the αpt by 5 to 8-fold. On the flip impact, Gao et al. (2005) found that the RCF(water) of PHN and pyrene with Chinese cabbage (flip = 0.0068) and ryegrass (flip = 0.0032) are largely proportional to root flip values. Similarly, the observed RCF(water) of di- and tri-chlorobenzenes with three crop roots (spinach, carrot, and radish) from farm sites conform largely to the order of root flip values (Zhang et al., 2005). The results demonstrate that plant flip and system settings have a major impact on RCF(water) for chemicals in roots with large flipKlip/fpw. However, as the flipKlip/fpw ratio becomes small, the influences of both plant flip and system settings should decline sharply or become unnoticeable.

Influences of field sample inhomogeneity

The field RCF(water) data are expected to be more variable due to spatially inhomogeneous fom or Com. This effect could be best illustrated by the field RCF(water) data with small flipKow (say, ≤5) because the expected variation in RCF(water) for systems having relatively uniform fom and Com in soil should be small according to the quasi-equilibrium model. In a field study of Mikes et al. (2009), the collected soil and radish samples of α-, β-, and γ-HCH, which have flipKow = 5.2–6.5, exhibit about a 10-fold scatter in RCF(water), with many (but not all) measured RCF(w) values (0.15–0.35) falling far below ~0.9 to yield very small αpt values (0.024–0.056). These anomalous data might have resulted from multiple causes, including highly non-uniform soil fom or chemical Com values, the combined errors in sample preparation and analyses, and/or other unknown causes. In contrast, the RCF(water) data reported by Zhang et al. (2005) for several chlorobenzenes on a few crop roots from soils (with flipKow = 2.5–35) at some farm sites display only 2 to 5-fold changes. As the field-sample homogeneity could vary widely, a potentially large variation in RCF(water) or αpt may occur.

Influences of chemical degradation and metabolism

Chemical degradation in soil rhizosphere and metabolism inside plants affect the plant contamination levels or RCF(water). The impact of either an abiotic breakdown or a microbial degradation in the rhizosphere on plant/crop uptake could be treated in terms of the reduced Com or Cpw value in Eq. (6), which mitigates the plant uptake level (Cpt) but exerts no direct effect on the RCF(water). In this case, the degraded species is viewed as a new chemical subject to the same partition-limited plant uptake. On the other hand, a metabolic process in a plant may or may not affect the Cpt or RCF(water) of the parent chemical depending on the situation. Although the current data is inadequate to fully settle this issue, one expects that only when the chemical metabolic rate far exceeds the rate of chemical uptake by plants will the metabolism have a strong impact on the RCF(water) or αpt value.

The above view is compatible with the observation of Trapp et al. (1990) on uptakes of atrazine (log Kow = 2.71), 1,2,4-trichlorobenzene (TCB) (log Kow = 3.98), and other chemicals in a soil by barley seedlings (flip = ~0.01) over one week of exposure. Here the log RCF(water) of atrazine and TCB are 0.80 and 1.28 and the αpt 1.0 and 0.20, respectively (see Table S2). Masses of atrazine and TCB metabolites in barley are 148% and 90%, respectively, of the parent chemicals. As seen, log RCF(water) values of atrazine and TCB fall close to the regressed line in Fig. 3 and the αpt values (≤1), especially 1.0 for atrazine, also score well with others in Fig. 4. This indicates that metabolism in plants does not necessarily affect the partition uptake of parent chemicals. Thus, for atrazine and TCB in barley (with flipKlip/fpw = 5.7 and 105, respectively), the metabolism seems to have no major effects on RCF(water) and/or αpt. For other compounds (e.g., hexachlorobenzene and DDT) having large flipKlip/fpw but insignificant metabolic breakdown in barley (Trapp et al., 1990), observed RCF(water) and αpt values are also compatible with those of other chemicals. Thus, unless the αpt values are exceptionally small and high levels of metabolites are detected, small αpt values for chemicals with high flipKlip/fpw could well be the result of slow chemical uptakes by plants alone.

Chemical uptakes by shoots and plant seedlings

With the observed efficiency of the quasi-equilibrium partition model for estimating the levels of chemicals in plant/crop roots, a brief account of the model applicability for shoot uptakes is in order. Although the current shoot-uptake studies with reported shoot flip values are limited, the shoot concentration factors, SCF(water), of many PAHs (Gao et al., 2005; Tao et al., 2009) based on their levels in soils and shoots and estimated Cpw are compatible with αpt ≤ 1 imposed by the model. Related SCF(water) and αpt data with plant/crop shoots are listed in Table S4, the SI. Observed log SCF(water) or αpt values for PAHs with shoots scatter widely when plotted against the shoot log flipKow (see Figs. S2 and S3, the SI). Overall, with a plant/crop, the SCF(water) for given PAHs are lower than the respective RCF(water) by 2–100 times such that no simple correlation stands out for estimating the SCF(water) with shoots.

On the other hand, it is encouraging to find that the quasi-equilibrium partition model (Eq. (15)) gives satisfactory estimates of the RCF(water) values with whole plant seedlings (Li et al., 2002; Su and Zhu, 2006; Su et al., 2009; Trapp et al., 1990). This outcome reflects that the total transport paths of chemicals from external water to small plants are relatively short. This observation adds values to the quasi-equilibrium partition model to estimate the levels of contamination in small food crops from soils or water solutions. However, more studies on latter systems are needed to increase the database for developing improved system-specific correlations.

Prospects on estimation of crop contamination

As exhibited in Figs. 1, 3, and 4, the partition-limited model (Eq. (7)) consolidates the water-plant and soil-plant RCF(water) data encompassing diverse system settings. The observed convergence of the RCF(water) values with soils and water stems from the use of soil Cpw in Eq. (1) as the chemical driving force for plant uptake. The other key parameter sharpening the RCF(water) estimation is the flipKow used for quantifying the lipophilic capacity of a specific chemical-plant pair. In essence, the highly improved correlation of Eq. (15) delineates the dependence of RCF(water) on the chemical hydrophilicity/lipophilicity and the system setting.

There is a special merit to estimate the RCF(water) for chemicals with log flipKow ≤ 2. With root flip being ~0.001–0.01, this log flipKow corresponds to log Kow ≤ 4–5, which covers a great many pesticides. In this range, the average αpt values with roots fall into 0.20–1.0; consequently, the variation of αpt with system settings is much reduced. Thus, once the ranges of αpt for chemicals with certain plant roots and system settings are determined, they could be used for estimating the RCF(water) values or chemical levels in roots at some chemical levels in soil or water under similar settings. In this practice, the finding that αpt is not sensitive to fom or Com for well-mixed soils (Beestman et al., 1969; Gao et al., 2005; Harris and Sans, 1967; Tao et al., 2009) facilitates the said RCF(water) estimation, with the more variable field-site samples excluded.

Conclusions

Analysis of the extensive soil-plant distribution data for a wide variety of nonionic organic chemicals shows clear evidence that the RCF(soil) based on Cs in a soil is an uncritical distribution index because it is a function of the SOM content. By adopting the chemical levels in soil pore water (Cpw), based on the estimated levels in SOM (Com), the resulting RCF(water) effectively overcomes the deficiency of RCF(soil). To further improve the analysis, a more quantitative lipophilic parameter (flipKow) is employed to account for the lipid effect on RCF(water). By these adjustments, the observed log RCF(water) from diverse soil-plant settings are eminently related to the associated log flipKow (R2 = 0.825, n = 368). This suggests that RCF(water) is an effective bioconcentration index and flipKow an essential parameter for the organic-chemical uptake by roots. The quasi-equilibrium values (αpt) found for chemicals under different system settings exhibit an anticipated trend that the ranges of αpt or RCF(water) at given flipKow are reduced at small flipKow values, with the field data excluded. This effect aids in estimation of the root contamination levels of chemicals (especially, pesticides) with log Kow ≤ 5. The intimate correlation of RCF(water) with flipKow should enhance the transport-fate modeling of nonionic contaminants in soil-plant systems. Information from this work offers a far more accurate prediction of the organic bioaccumulation by plant/crop roots and thus a more accurate risk assessment of the human exposure.