Mineralization of 17β-estradiol in 36 surface soils from Alberta, Canada

Recent column studies suggest that pesticide fate models could be used to estimate the fate of estrogens in soil. Estrogens are detected in livestock manure which is used as a nutrient source on agricultural land. This is the first study to examine estrogen mineralization in a wide range of agricultural soils at the regional-scale. Soil samples were collected from upper and lower landscape positions of 18 agricultural fields in an area spanning 49-60 degrees N longitude and 310-120 degrees W latitude and these samples were used to determine 17 beta-estradiol mineralization parameters in microcosm experiments. Maximum 17 beta-estradiol mineralization (Max) ranged from 5.8% to 19.2% and was, on average, significantly (P < 0.05) less than the 47.9-61.9% range measured in the same soils for the widely-used herbicide 2,4-D (2,4-dichlorophenoxyacetic acid). Maximum 17 beta-estradiol mineralization was positively correlated to Kd-17 beta-estradiol (r = 0.62, P < 0.001) while 2,4-D maximum mineralization was negatively correlated to Kd-2,4-D (r = -0.52, P < 0.01), even though 17 beta-estradiol and 2,4-D sorption parameters (Kd) were positively correlated (r = 0.62, P < 0.001), and both Kd-17 beta-estradiol and Kd-2,4-D values were significantly positively correlated to SOC (r = 0.71, P < 0.001; and r = 0.67, P < 0.001, respectively). Hence, the mineralization of 2,4-D decreases as its sorption to soils increases while the mineralization of 17 beta-estradiol increases as its sorption to soil increases. This suggests that some steps in the 17 beta-estradiol mineralization process are occurring in the sorbed phase. Equations to predict 17 beta-estradiol and 2,4-D sorption and mineralization parameters were established using Partial Least Squares regression. Significant models for mineralization (r(2) from 0.42 to 0.56) had lower r(2) than significant sorption models (r(2) from 0.78 to 0.85). Given the poor results of the mineralization regression models, we conclude that probability density functions, rather than regression models, are likely to be more useful for describing pesticide or estrogen mineralization parameters at the regional scale. Based on our findings, agri-environmental policy-analysis in Alberta should use the log-logistic probability density function to describe the mineralization rate (k) of either 17 beta-estradiol or 2,4-D at the large-scale. Max-17 beta-estradiol at the large-scale is best described by the extreme values probability density function and Max-2,4-D by the triangular probability density function. (c) 2010 Elsevier B.V. All rights reserved.