Statistical approach to assess radon-222 long-range atmospheric transport modelling and its associated gamma dose rate peaks

International audienceThere is a need for a validation framework for long-range atmospheric transport modelling dedicated to radionuclides. For distances greater than 50 km, the modelling of radionuclide deposition and ambient gamma dose rate evaluation are particularly difficult to validate, since it has been mainly only observed after the accidents of Chernobyl and Fukushima. There is however a natural wet deposition phenomenon leading to numerous well-observed gamma dose rate events: the scavenging of radon-222 progeny by rain. Radon-222 exhalation from the soil to the atmosphere, its decay, its progeny, its own transport, the transport of its progeny, their deposition, and the consequent ambient gamma dose rate are then modelled at the European scale. This whole atmospheric radon model from soil (exhalation) to soil (deposition) needs to be validated by comparison with observations. The biggest benefit of this case study is the number of events that serve as a comparison. For a statistical evaluation of the performance of the model, we compared its results with gamma dose rate observations over a period of two years, gathering more than 15 000 peaks greater than 10 nSv h−1 above the background radiation. Two sets of metrics were used to assess the agreement between the model and observations: on a case by case basis (peak to peak) and continuously (whole time series of gamma dose rates and air concentrations). Particular attention was paid to defining the metrics in order to remove the background radiation level and to exclude outlier stations. We found that 48 % of the gamma dose rate peaks are well modelled, a fraction of which can rise up to 89 % by being more tolerant with the success criteria. The model has proven to be of the correct magnitude, with room for substantial improvement. Overall, the modelling shows better recall than precision: i.e. a tendency to produce more false positives than false negatives. It is also less effective in reproducing the highest peaks. Exhalation, vertical mixing and deposition have been identified as the three main features which could improve this model. Now validated, with all its limitations, the atmospheric radon model may serve for its primary purpose, the validation of atmospheric transport modelling and its input data. It also may serve as a framework to test any exhalation model on a national or continental scale. Moreover, it is useful to learn how to properly use the data of an ambient gamma dose rate network, and how to compare this data to modelled data. Finally, some interesting features concerning the assessment of outdoor concentrations of radon-222 became apparent