Modelling the effects of bioturbation on the re-distribution of ¹³⁷Cs in an undisturbed grassland soil

Under favourable conditions, soil ingestion by earthworm populations can be equivalent to approximately 5-10% of the topsoil mass per year. This suggests that for contaminants that are strongly bound to soil, earthworm 'bioturbation' may be a more important transport mechanism than water-borne advection dispersion. It is therefore quite surprising that few modelling studies to date have explicitly considered the effects of biological processes on contaminant transport in soil. In this study, we present a general model that incorporates the effects of both 'local' and 'non-local' biological mixing into the framework of the standard physical (advective-dispersive) transport model. The model is tested against measurements of the redistribution of caesium-137 (¹³⁷Cs) derived from the Chernobyl accident, in a grassland soil during 21 years after fallout. Three model parameters related to biological transport were calibrated within ranges defined by measured data and literature information on earthworm biomasses and feeding rates. Other parameters such as decay half-life and sorption constant were set to known or measured values. A physical advective-dispersive transport model based on measured sorption strongly underestimated the downward displacement of ¹³⁷Cs. A dye-tracing experiment suggested the occurrence of physical non-equilibrium transport in soil macropores, but this was inadequate to explain the extent of the deep penetration of ¹³⁷Cs observed at the site. A simple bio-diffusion model representing 'local' mixing worked reasonably well, but failed to reproduce the deep penetration of Cs as well as a dilution observed close to the soil surface. A comprehensive model including physical advective-dispersive transport, and both 'local' and 'non-local' mixing caused by the activities of both endogeic and anecic earthworms, gave an excellent match to the measured depth profiles of ¹³⁷Cs, with predictions mostly lying within confidence intervals for the means of measured data and model efficiencies exceeding 0.9 on all sampling occasions but the first.