Measurements at only 117 of 121 grid points were used because standing water in the southeast corner of the field prevented access to three locations and the fourth measurement was lost. The mean Kd for atrazine was 4.94 L kg-1 (Table 14.1) and ranged from 1.74 to 10.92 L kg-1 with the sample population distribution being positively skewed (mean > median). SOC averaged 0.0280 kg kg-1 and varied between 0.0123 and 0.0556 kg kg-1, and the distribution was also positively skewed. Both Kd and SOC were better described by lognormal distributions than normal distributions as determined by the Kolmogorov-Smirnov test and the D'Agostino-Pearson test (Cambardella et al., 1994). ECa measurements ranged from 20.4 to 65.4 mS-1 and averaged 41.0 mS-1. These values are relative, however, and are not direct measures of the true soil electrical conductivity (Lesch et al., 1992). ECa data were not described well by either normal or lognormal distributions, although the relative difference between the mean and median was less for the untransformed data.

The spatial distribution of each parameter over the intensive-grid area is shown in Figure 14.2. None of the parameters were randomly distributed across the area but were instead grouped into areas of like values. Moran' s I statistic for Kd, SOC, and ECa was 0.52, 0.74, and 0.59, respectively, indicating substantial spatial clustering. Each parameter showed lower values in areas of the field mapped as well-drained Clarion loam and higher values in the area mapped as very poorly drained Okoboji mucky silt loam. This agrees with our hypothesis that all three parameters should be correlated with drainage class.

In addition to similar spatial patterns, the spatial structures of the three parameters were almost identical as determined by similar correlograms (not shown). The correlograms indicated a spatial dependence between measurements for each parameter to distances of about 80 m. The similar spatial distribution leads to a significant simple correlation statistic between each pair with the correlation between ECa and SOC being the strongest (Table 14.1). Given the similar spatial patterns, spatial structure, and positive correlation between the parameters, it appears likely that Kd can be successfully estimated from either SOC or ECa. Koc was calculated by regressing the Kd values versus the SOC values using linear least sum-of-squares. The resulting value for Koc was 0.171 L kg-1 with a standard error of regression of 1.42 L kg-1.

An analogous regression was performed between Kd and ECa. However, because the Kd values were lognormally distributed and the ECa values were better described by a normal distribution, the expression equivalent to Koc is ln Kd = b ECa or Kd = exp (b ECa), where b was found by nonlinear

0 0

Post a comment