Evaluating summarised radionuclide concentration ratio datasets for wildlife

Concentration ratios (CR(wo-media)) are used in most radioecological models to predict whole-body radionuclide activity concentrations in wildlife from those in environmental media. This simplistic approach amalgamates the various factors influencing transfer within a single generic value and, as a result, comparisons of model predictions with site-specific measurements can vary by orders of magnitude. To improve model predictions, the development of 'condition-specific' CR(wo-media) values has been proposed (e.g. for a specific habitat). However, the underlying datasets for most CR(wo-media) value databases, such as the wildlife transfer database (WTD) developed within the IAEA EMRAS II programme, include summarised data. This presents challenges for the calculation and subsequent statistical evaluation of condition-specific CR(wo-media) values. A further complication is the common use of arithmetic summary statistics to summarise data in source references, even though CR(wo-media) values generally tend towards a lognormal distribution and should, therefore, be summarised using geometric statistics. In this paper, we propose a statistically-defensible and robust method for reconstructing underlying datasets to calculate condition-specific CR(wo-media) values from summarised data and deriving geometric summary statistics. This method is applied to terrestrial datasets from the WTD. Statistically significant differences in sub-category CR(wo-media) values (e.g. mammals categorised by feeding strategy) were identified, which may justify the use of these CR(wo-media) values for specific assessment contexts. However, biases and limitations within the underlying datasets of the WTD explain some of these differences. Given the uncertainty in the summarised CR(wo-media) values, we suggest that the CR(wo-media) approach to estimating transfer is used with caution above screening-level assessments.