Uncertainty analysis of partition coefficients in a radionuclide transport model

Models of transport and pollutant dispersion are routinely used to define risks posed by contaminants present in aquatic environments. Modelers frequently utilize a simple parameter, known as a partition coefficient, to apportion contaminant concentrations between dissolved and particulate phases. But the inherent uncertainty in the master variables for partition coefficients (distribution coefficient (Kd); particulate matter concentration) is also a source of uncertainty in model results. We demonstrate an approach that can be used to quantify this source of uncertainty and apply it to an investigation of the transport of radionuclides away from a radioactive waste site in a shallow Arctic Bay. Our investigation considers the radionuclides, americium, plutonium, europium, ruthenium, cobalt, cesium, and strontium which exhibit a broad range of affinities for marine particles. Probability distribution functions were constructed for radionuclide distribution coefficients (Kd) and particulate matter concentrations using Beta–Pert functions. Simulations using Latin Hypercube sampling were performed on the functions of Kd and particulate matter to define probability functions of partition coefficients for each radionuclide. These distributions were used in a hydrodynamic model of Abrosimov Bay to predict ranges of radionuclide concentrations for different atmospheric forcing events. From this investigation, partition coefficients values (PCw) exhibit the highest uncertainty for radionuclides having mid-range Kd values (Kd=1–30 m3/kg). When applied to model simulations of radionuclide transport, concentrations of less particle-reactive radionuclides, strontium (Kd=0.01 m3/kg) and cesium (Kd=0.3 m3/kg), are most sensitive to variations in wind direction and hence to water exchange rates. Whereas europium (Kd=100 m3/kg), ruthenium (Kd=30 m3/kg) and cobalt (Kd=1 m3/kg) concentrations exhibit large variations in response to different combinations of partition coefficient values and wind events. Even high Kd radionuclides such as americium (Kd=1000 m3/kg) exhibit large uncertainties in dissolved concentration when simulating low sediment concentrations for the bay. Generalizing these findings, the uncertainty in the master variables used to calculate partition coefficients is significant for all but the most soluble contaminants. The method described in this paper can be a useful tool in radionuclide transport modeling for quantifying this uncertainty.