| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Chemistry Dep., The Royal Veterinary and Agricultural University, Thorvaldsensvej 40, DK-1871 Frederiksberg C, Copenhagen, Denmark
b Informatics and Mathematical Modelling DTU, Building 321, Technical University of Denmark, DK 2800 Lyngby, Denmark
c Environment & Resources DTU, Building 115, Technical University of Denmark, DK 2800 Lyngby, Denmark
* Corresponding author (peho{at}kvl.dk)
Received for publication November 14, 2001.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Abbreviations: CBD, citrate–bicarbonate–dithionite • CEC, cation exchange capacity • HIM, hydroxy interlayered clay minerals • TEA, total element analysis • XRD, X-ray diffraction
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Several studies have correlated observed Kd values for Cd with soil characteristics. Anderson and Christensen (1988) found that pH was the most influential factor in controlling the distribution of Cd in soils. Christensen (1989) correlated Cd Kd values determined for 63 soil samples (21 agricultural profiles represented as three depths) to soil parameters such as pH, texture, humus, and CEC and components such as extractable iron, manganese, and aluminum oxyhydroxides. Equilibrium pH was the dominant parameter in the stepwise regression analysis accounting for 72% of the variation in Kd. Including the soil humus content improved the correlation coefficient by a few percentage points. No significant improvement in correlation coefficient was obtained by introducing further soil parameters or components. Sauvé et al. (2000) recently reviewed and compiled data from more than 70 studies and found by multiple linear regression analysis that approximately 50% of the overall variation in the Kd values for Cd could be explained by variations in solute pH. However, including soil organic matter as a second component significantly improved the regression (R2 = 0.61). Other studies have also identified pH (e.g., Buchter et al., 1989; Gerritse and van Driel, 1984; Naidu et al., 1994; Sánchez-Camazano et al., 1998) and in some cases organic matter (e.g., Boekhold and van der Zee, 1992; Gerritse and van Driel, 1984; Gray et al., 1998) and cation exchange capacity (e.g., Sánchez-Camazano et al., 1998) as the soil parameters providing the strongest correlation with Kd values for Cd.
pH may also correlate strongly with soil components such as clay minerals, oxides, or organic matter and hence mask their contribution to the Cd sorption. Studies involving a single sorbent have been conducted on synthesized components, pure minerals, and fractions separated from whole soils (e.g., organic matter, clay fractions, and clay minerals) revealing that many components have the ability to sorb Cd. However, this does not reveal which components govern Cd sorption in whole soils, where various components will interact (e.g., as coatings on other components) and compete for Cd sorption.
The purpose of this study is to determine which soil characteristics may be important in controlling Cd sorption. Soil characteristics are soil parameters, reactive soil components, and clay minerals quantified in the soils. To eliminate the effect of pH, the Kd values in different soils were determined at two preset pH values and the statistical analysis performed for each data set.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
The carbon content was determined gravimetrically after dry combustion in a LECO (St. Joseph, MI) apparatus according to the method described in Tabatabai and Bremner (1970). The CEC of the soil samples was determined by the ammonium acetate exchange method (Rhoades, 1982). The particle size distribution was determined by sieving and pipette methods after removal of organic matter with 30% hydrogen peroxide and dispersion in 0.002 M sodium pyrophosphate (Andreasen, 1939). The pyrophosphate-extractable Fe and Al determination was performed according to the methods described by McKeague (1967), the acid ammonium oxalate–extractable Fe, Al, and Si determined as described by Schwertmann (1964), and the citrate–bicarbonate–dithionite (CBD)-extractable Fe, Al, and Si determined according to the method described by Mehra and Jackson (1960).
The clay fraction (<2 µm) was separated from the silt and sand fractions using settling in water and a particle size centrifuge (Slater and Cohen, 1962) after removal of organic matter by means of 30% hydrogen peroxide. Following separation the clay was flocculated by addition of different salt solutions and air-dried. The so-called free aluminum and iron oxides were removed from the clay by CBD treatment (Mehra and Jackson, 1960). The composition of the CBD-treated clay fractions was estimated by the combination of methods described by Møberg et al. (1988). In summary, the estimation is based on total element analysis (TEA), performed on NH4–saturated clay as described by Bernas (1968), CEC determination (Rhoades, 1982), and X-ray diffraction (XRD) analysis. The mineralogy was assessed by XRD, which was performed on a Philips (Eindhoven, the Netherlands) 1050 vertical diffractometer with CoK
radiation using unoriented mounts on Mg-saturated clay and oriented mounts prepared from CBD-treated samples saturated with Mg (at room temperature and following glycerol treatment) and K (at room temperature and following heating at 550°C for 2 h). The XRD analysis showed the presence of the following minerals: quartz, feldspars (microcline and plagioclase), gibbsite, chlorite, kaolinite, illite, smectite, and HIM (i.e., hydroxy-interlayered smectite and vermiculite). Ordinary vermiculite was also included in HIM due to difficulties in distinguishing between these minerals. The calculation of the mineral composition of the clay fraction was based on the following assumptions: (i) All the Ca and Na determined by TEA were considered to belong to plagioclase (Na–Ca–feldspar). (ii) The plagioclase to microcline (K–feldspar) ratio was estimated from the intensities of the main XRD peak of each of these minerals. (iii) The rest of the K was considered to belong to illite, if this mineral was identified. (iv) After allocating the other necessary elements and CEC to these minerals the remaining Mg and Fe were distributed between the other layered silicate clay minerals (chlorite, HIM, smectite) found by XRD analysis taking into consideration the CEC. (v) After allocation of necessary elements, the rest of the Al was allocated to kaolinite, if present. (vi) Finally, the Si left over was allocated to quartz if this mineral was found by the XRD analysis. As all the calculations were based on percentage distribution the percentage sum of the minerals should add up to approximately 100%. Furthermore, the distributed minerals should account for the determined CEC. In these calculations, allocations of elements to the various minerals were based on assumptions about the elemental composition of the various minerals. Mineral composition can, however, exhibit considerable variation especially for chlorite, HIM, illite, and smectite, whereas the composition of the feldspars, gibbsite, kaolinite, and quartz was rather fixed.
Sorption Experiments
Cadmium sorption experiments were performed on all 49 soil samples. In addition, duplicate experiments were undertaken for three soil samples. For each soil sample, Cd distribution coefficients (Kd) were determined at two fixed pH values: 5.3 and 6.7.
Each sorption experiment involved 2 g of soil and 50 mL of salt matrix (10-3 M CaCl2) equilibrating in a 100-mL polyethylene bottle for 24 h. pH was adjusted to 5.3 or 6.7 with small amounts of HNO3 and NaOH. This adjustment was repeated until pH was approximately constant. Cadmium was added from a stock solution [Cd(NO3)2]. The amount of Cd added was adjusted to each individual soil sample, based on preliminary experiments, to obtain equilibrium solution concentrations in an environmentally relevant low range. At equilibrium the Cd solution concentration was between 0.08 and 17 µg L-1 at the lowest Cd loading and between 0.16 and 34 µg L-1 at the highest loading of Cd. After addition of Cd, the bottles were agitated for another 24 h and pH was adjusted to the fixed value. Before separation of solid and solution, the pH was within ±0.03 pH units from the fixed value. Solids and solution were separated by centrifugation (equivalent particle separation diameter: 0.2 µm) and the solution was preserved with acid (HNO3) and stored for analysis.
The concentration of Cd sorbed on the soil was calculated from the mass balance as the difference between Cd added and Cd in the solution after equilibration. Each Kd value was obtained by linear regression based on two individual experimental observations representing different Cd concentrations on the sorption isotherm and on (0,0) of the isotherm. Preliminary experiments (data not shown) involving four soils had shown that the isotherms were linear and that the original Cd content of the soil samples was negligible.
Cadmium Analysis
The acidified solution samples were analyzed for Cd by graphite-furnace atomic absorption spectroscopy (PerkinElmer [Wellesley, MA] 5000, deuterium background correction, HGA 400 graphite furnace, AS-1 automatic sample injection system) after liquid–liquid extraction (1.0 Na-diethyldithiocarbamate, trihydrate in 4-methylpentan-2-one). This method, as used by, for example, Holm et al. (1995), reduces the chemical interference and provides low detection limits.
Statistical Analysis
Statistical analysis was performed on 98 Kd values (49 samples). The Kd values were not normally distributed (shown from the residuals of the relevant models by a 
2 test), but transforming the Kd values by the logarithm (base 10) gave reasonable results. Therefore, the statistical analyses were based on the Kd values transformed to the logarithm (base 10) of this value, in the following denoted by log(Kd). To make an interpretable multiplicative model, all soil characteristic variables (except Kd) were raised by the value 1.0 prior to the logarithm (base 10) transformation. This makes all transformed values positive and hence simplifies the interpretation of correlation coefficients.
The correlation matrix for all the variables (Table 4) shows that many of the variables are highly correlated, suggesting grouping of the variables by principal component analysis. The grouping was partly based on the correlation matrix and partly on knowledge of soil characteristics and constituents. The groups were used as first variables in the regression analysis together with some ungrouped variables. When a reasonable model was identified, each group was substituted by the single variable from the group that gave the overall best fit for the final model. It should be noted that, by analyzing a data set in this way, more models, depending on the actual procedure, could be identified yielding equally good representation of the Kd values. As a part of the procedure, it is thus necessary to incorporate knowledge of soil characteristics and the chemical constituents. The statistical analysis was performed by the use of SAS (SAS Institute, 1996).
|
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
The pH of soil MJ1 is somewhat lower than the pH of the other soils and the pH increases with increasing depth in soil MJ1, while the other soils exhibit decreasing or almost constant pH at increasing depth. This is due to the fact that MJ1 is a forest soil, while the other soils are agricultural soils that are limed regularly.
The organic carbon contents cover a wide range from 0.02 to 6.0% (Table 1). The carbon content generally decreases with increasing soil depth because the organic carbon originates from dead organic matter such as crop residues, weeds and grasses added to the soil surface, and roots concentrated in the upper part of the soil. However, in two of the soils, MJ1 and MJ4, the maximum C content occurs in the Bh horizon, because these soils are Spodosols formed by translocation of humus together with aluminum and iron from A and E horizons to the B horizon (Borggaard, 2002). Due to this translocation, MJ1 and MJ4 as well as SJ2 and SJ3, which are also podzolized soils, have the greatest amounts of pyrophosphate- and oxalate-extractable aluminum and iron in the subsoil. Pyrophosphate-extractable iron occurs as organic complexes in soil, while acidified oxalate extracts iron and aluminum in organic complexes and poorly ordered oxides and silicates. Citrate–bicarbonate–dithionite also extracts crystallized iron oxides (Borggaard, 1990), whereas CBD-extractable aluminum presumably includes aluminum present in organic complexes and poorly ordered oxides and silicates as well as aluminum trapped by isomorphic substitution in crystalline iron oxides. The similarity in amounts of aluminum extracted from the eight soils by the three extractants (Table 1) is therefore an indication of little aluminum substitution for iron in these soils. On the other hand, all soils contain crystallized iron oxides as seen by the difference between CBD-extractable and oxalate-extractable iron, although the contents are small except in the clay soil (SJ5). The amount of oxalate- and CBD-extractable silica originating from adsorbed and poorly ordered silicates such as allophane and imogolite (Borggaard, 2002) is low, indicating the absence of allophane, imogolite, and other poorly ordered aluminum silicates in these soils.
The particle size distribution exhibits substantial variation in clay, silt, and sand contents ranging from 0.2 to 58% (clay), 0.2 to 38% (silt), and 4 to 99% (sand). However, most samples have clay and silt contents of less than 5%, reflecting their formation on sandy parent materials. With a clay content of up to 58% and as little as 4% sand, the SJ5 soil formed on very clay-rich till deviates from the general pattern. Moreover, the subsoil in MJ4 is somewhat richer in clay than the upper part of the soil.
The CEC is largely determined by the organic carbon and clay contents. So, the highest CECs are found in soil samples enriched in organic carbon and clay, as in SJ5 samples with high clay contents and some of the B horizons of the Spodosols (MJ1, MJ4, SJ2) with rather high contents of illuvial organic matter (Table 1).
In the interpretation of the composition of the clay fractions shown in Table 2 it must be recognized that the values are semiquantitative estimates, as stated earlier in this paper. Despite the uncertainties, the clay fractions of the eight soils have a composition comprising several primary minerals (microcline, plagioclase, quartz) and secondary minerals (chlorite, gibbsite, HIM, illite, kaolinite, smectite) as also found in most other Danish soils (Møberg et al., 1988). Illite, kaolinite, and quartz are the dominant minerals in all the soils. The smectite content is substantial in the soils (SJ2, SJ3, SJ5) and in the upper horizons of MJ4, whereas this mineral could not be detected in the other soil samples. All samples contain minor amounts of feldspars with microcline as the most abundant one. The gibbsite contents are low, especially in the upper part (A and E horizons) of the podzolized soils (MJ1, MJ4, SJ2, SJ3) (data on vertical distribution not shown). Stability and mode of formation may explain the presence of HIM in even the most strongly weathered Spodosols investigated (Table 2).
|
The Kd values for Cd in the different horizons of the eight soil profiles are presented in Table 3. As expected, all Kd values are greater at pH 6.7 than at pH 5.3. Generally, the Kd values are greater in the upper horizons of the soil profiles at both pH levels than in the lower horizons. The measured Kd values in the A horizons ranged from 70 to 850 L kg-1 at pH 5.3 and 460 to 2940 L kg-1 at pH 6.7. These values are in the same range as the Kd values reported for other topsoil samples by Christensen (1989). Christensen (1989) found likewise that the samples from 0.5- to 1.0-m depths tended to show somewhat lower Kd values than observed in the topsoil. This study differs from other studies by fixing the soil pH at two specific levels. Despite this, the Kd values still vary substantially from soil to soil, indicating that factors other than pH are important in controlling Kd.
|
Two regression analyses were made: One for each pH level. For pH level 5.3 the significant parameters for describing the Kd values were: percent C, CEC, and gibbsite. For pH level 6.7 the significant parameters were percent C and pyrophosphate-extractable Fe. The R2 values were 0.78 (pH 5.3) and 0.55 (pH 6.7). The analysis gave the following model regression equations at the fixed pH levels:
 |
 |
 |
 |
The equations are multiplicative expressing that a parameter value of zero will correspond to a multiplication by a factor of 1.0. Given identical values for FePyr, the regression equation (pH 6.7) predicts that if percent C changes from 0.5 to 1.0 the change in Kd corresponds to a factor of 2.5.
From the stepwise regression it is recognized that multicolinarity among the explaining variables makes it impossible to find the single "best" model. Many models will be comparable with regard to R2 values. Therefore, interpretation should be made with precautions.
The results of the statistical analysis show that when the effect of pH is eliminated by fixation of pH, soil characteristics such as clay content, clay minerals, and most reactive parameters are of no statistical importance for the variation in Cd soil Kd values. The only parameters in addition to pH are the percent C and at low pH the CEC and gibbsite content. At high pH the pyrophosphate-extractable Fe plays a role.
The finding that soil organic matter (represented by the percent C) is a significant component is in accordance with several previous studies that have identified it as one of the components controlling the distribution of Cd in soils, for example, Boekhold and van der Zee (1992), Christensen (1989), and Sauvé et al. (2000).
Although the number of sorption sites occupied by Cd is low compared with the cation exchange capacity of the soil, CEC is a statistically significant parameter, supposedly because CEC is correlated with the humus content. The humus is a variable charge component with a relatively protonized surface at pH 5.3.
The gibbsite content is negatively correlated with Kd, contributing negatively to the log(Kd) value at pH 5.3. The gibbsite content is zero or very low in all soil samples and in those samples where it is present in highest contents (up to 8% of the clay fraction) its effect on the Kd value seems to be detectable. Although statistically significant at pH 5.3, the content of gibbsite in soil samples is considered to be only of minor importance in describing the variation in Cd Kd values.
Pyrophosphate-extractable Fe is significantly correlated with Kd at pH 6.7. This parameter is a measure of Fe in complexes with organic matter, or maybe more correctly in Fe oxide–organic matter associates (Borggaard, 2002). At increasing pH the negative charge, and hence affinity for metal ions like Cd2+, of these associates will increase substantially as both Fe oxides and organic matter become more negative at higher pH. This will explain why pyrophosphate-extractable Fe is statistically significant at the higher pH.
Except for gibbsite discussed above, none of the clay mineral fractions were important in describing the variation in Cd distribution coefficients. Based on the literature, the importance of the different types of clay minerals can be observed only in experimental systems with pure types of clay minerals, but apparently the significance of clay minerals cannot be observed in whole soil samples containing a relatively low percentage of clay, which is composed of mixtures of different clay minerals.
The uncertainty of the regression equations derived is illustrated in Fig. 1 showing the predicted values of log(Kd) against the measured values at the two fixed pH values. The derived models at two different fixed pH values is here used to predict Cd Kd values at these two actual pH values as shown in Fig. 1. However, as the models are developed for two fixed pH levels, the models are not appropriate for predictive purposes as Cd Kd values typically are required for a range of pH values.
|
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Crop Science | |||
| Vadose Zone Journal | Journal of Plant Registrations | ||||
| Journal of Natural Resources and Life Sciences Education |
Soil Science Society of America Journal |
