MODELING RESULTS

MODELING RESULTS
Based on preliminary analyses, regressions of plant concentrations versus soil concentrations using salt data differed significantly from those using field data. Significance values (p) were 0.035 for arsenic, 7.9 x 10-15 for cadmium, 4.5 x 10-5 for copper, 0.0017 for lead, 0.0059 for mercury, 0.00011 for nickel, 7.2 x 10-29 for selenium, and 0.0013 for zinc. For some of the chemicals (arsenic and nickel), the salt uptake data were within the 95% prediction limit of the field data regressions (Appendix C). However, or most chemicals, several data points were outside of these bounds. Most concentrations of selenium in plants when the source of selenium was selenate or selenite were higher than most concentrations of selenium in plants in the field studies (Appendix C, Fig. C.7). For some chemicals, such as arsenic, cadmium, and zinc, the plant concentrations associated with salts additions were comparable to the highest plant concentrations from the field dataset. For other chemicals, such as mercury, the range of
soil concentrations in the salts dataset was simply too narrow to give a good regression line (Appendix C, Fig. C.5). However, because for some chemicals, salts-amended soils were generally associated with higher chemical concentrations in plants than chemicals in field soils, the decision was made to exclude salts data from the models for uptake of all chemicals by plants. All results below are for field data only. Soil-plant regression models and uptake factors were developed for eight inorganic chemicals: arsenic, cadmium, copper, lead, mercury, nickel, selenium, and zinc (Figs. 1 to 8). In the initial dataset with salts data excluded, the number of observations ranged from 99 for arsenic to 164 for zinc (Table
1). The number of studies incorporated in the models ranged from seven for nickel to twenty for zinc. Six of eight distributions of uptake factors fit a lognormal distribution more closely than a normal distribution, though only the distribution of uptake factors for arsenic, lead, selenium, and zinc fit the lognormal form well (Table 1). Median uptake factors for all chemicals were less than one; however, the maximum uptake factor for all chemicals exceeded one. The distributions of uptake factors for the eight chemicals spanned at least two orders of magnitude; e.g., for copper the range of uptake factors was less than three orders of magnitude and for arsenic the range was greater than five orders of magnitude. An example of the cumulative distribution of uptake factors for selenium is presented in Fig. 9. [Note: the mean and standard deviation of the natural-log-transformed uptake factors are presented as parameters for describing the uptake factor distributions for chemicals where the distribution is lognormal .Whereas these untransformed uptake factors are best fit by a lognormal distribution, the natural-logtransformed uptake factors are normally distributed. These parameters may be used in two ways. They may be applied to normal distribution functions in Monte Carlo simulation software, however the output from the sampling from this distribution must be back-transformed. Alternatively, the parameters may be incorporated into the LOGNORM2 function in the @RISK1 Monte Carlo simulation software (Palisade Corp. 1994b). Use of the LOGNORM2 function requires no back-transformation. Comparable results are obtained using either approach.