Sensitivity and uncertainty analysis of a model computing radionuclides transfers in fluvial ecosystems (CASTEAUR): application to 137Cs accumulation in chubs

Sensitivity and uncertainty analysis were conducted on a stretch of the Rhone River (France) with CASTEAUR, a model computing radionuclides transfers to abiotic and biotic components of fluvial ecosystems. The sensitivity of 137Cs accumulation in omnivorous fish was analysed by constructing linear models between this output variable and some input parameters sets constructed using the Latin Hypercube sampling technique. When sedimentation occurred, the fish feeding ratio and the river suspended matter load explained the major part of the uncertainty associated to 137Cs accumulation in fish. Results of an uncertainty analysis were also confronted to the 137Cs accumulation measured in chubs (Leuciscus cephalus L.) of the Rhone River, downstream from a nuclear reprocessing plant. The model predictions and their associated uncertainties enclosed the observed radioactivity in fish, but only when suspended solids deposition was included in the model. We conclude from these observations that the sedimentary dynamics and the fish feeding habits should be properly characterized when assessing the impact of 137Cs effluents in rivers using mathematical simulation models such as CASTEAUR.