Sorption of four estrogens by surface soils from 41 cultivated fields in Alberta, Canada

Estrogenic compounds in livestock manure are also present in soils because manure is land applied on account of its value as an important nutrient source in agricultural production. This is the first study to compare the sorption of 17β-estradiol, estrone, estriol and equol in a wide range of soils. Specifically, for each of these four estrogens, the soil sorption coefficient (Kd) and the sorption coefficient per unit organic carbon (Koc) were determined in 121 surface soils (0–15 cm) obtained from upper-slopes, mid-slopes, and lower-slopes in agricultural fields across seven ecoregions in the Province of Alberta, Canada. Soil organic carbon content (SOC), texture, pH, carbonate content, electrical conductivity and cation exchange capacity were also determined on the air-dried and sieved (2 mm) soil. Both Kd and Koc values significantly increased in the order of Kd-estriol (23 L kg− 1 soil and 1059 L kg− 1, respectively) = Kd-17β-estradiol (23 L kg− 1 soil and 1082 L kg− 1, respectively) < Kd-estrone (33 L kg− 1 soil and 1557 L kg− 1, respectively) < Kd-equol (42 L kg− 1 soil and 2080 L kg− 1, respectively). For each estrogen, SOC was the strongest significant factor explaining variations in Kd values among soils and particularly small Kd values were observed when SOC was below a threshold value of 10 g C kg− 1. Regardless of the estrogen, the Mixed Grassland ecoregion characterized by Brown Chernozem soils, as well as upper-slopes in general, demonstrated lesser Kd values because of reduced SOC. However, regardless of the estrogen, the soils with reduced SOC displayed greater Koc values. Equations to predict 17β-estradiol, estriol, estrone or equol Kd or Koc values at the regional level were established using either soil properties (Partial Least Squares (PLS) regression) or one single estrogen (ordinary least squares regression). Regardless of the estrogen or the regression used, the strength of the prediction model, as determined by the coefficient of determination (r2) and other factors, was always better for Kd than Koc values. Regardless of the regression used, the r2 of the prediction models exceeded 0.70 for Kd-17β-estradiol and Kd-estriol, but r2 was below 0.52 for Kd-equol. For both Kd and Koc values, the prediction using soil properties (ranging from a r2 of 0.51 to 0.87 for Kd and from a r2 of 0.32 to 0.44 for Koc) always provided better prediction models than using a single estrogen (ranging from a r2 of 0.38 to 0.71 for Kd and from a r2 of 0.18 to 0.40 for Koc). We conclude that data on basic soil properties are good tools for estimating Kd values of 17β-estradiol, estrone and estriol in western Canadian soils. Additional studies are required to seek better prediction models at the regional scale for estimating Kd-equol and for estimating Koc values of estrogens, particularly because such information could be important for agri-environmental policy analyses in Canada and elsewhere.