EGU21-9081
https://doi.org/10.5194/egusphere-egu21-9081
EGU General Assembly 2021
© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

Statistical approach to improve Radon-222 long-range atmospheric transport modelling

Arnaud Quérel1, Khadija Meddouni1, Denis Quélo1, Thierry Doursout1, and Sonia Chuzel2
Arnaud Quérel et al.
  • 1IRSN, Fontenay-aux-Roses cedex, France (arnaud.querel@irsn.fr)
  • 2IRSN, Le Vesinet, France

The IRSN operates a framework capable of modelling the occurrence of gamma dose rate peaks due to Radon-222 progeny scavenged by precipitations. This framework includes a Radon-222 exhalation flux and meteorological data as inputs to a long range atmospheric transport model (ldx), and specific post-processes for reporting or performance analysis. Ldx is used in nuclear emergency preparedness and response, and it is applied to Radon-222 and its progeny to simulate both their air concentration and deposition onto the ground. This framework is successful at forecasting, in timing and intensity, around half of the peaks actually observed by the IRSN radiation monitoring stations over France. Understanding and analysing the failures is the starting point to improve the modelling.

For a statistical evaluation of the framework performance, we confronted its results to observations of gamma dose rates over a period of six months gathering more than 12,000 peaks. We used two sets of metrics to assess the agreement between model and observations: punctually (peak by peak) and continuously (whole six months’ time series of gamma dose rate and air concentration). We also performed statistical significance tests to identify the influence of some input parameters on the results.

We found that considering a factor 5 instead of a factor 2 between observed and simulated peak values increases the percentage of successfully forecasted peaks from around 50% to above 90%, whereas increasing the permissible time lag between the two has no such effect. Overall, the model shows better recall than precision: i.e. a tendency to produce more false positives than false negatives. ANOVA tests did not point out any performance difference across factors such as land cover or time of the day. Using weather radar measurements for precipitation instead of meteorological model data also improves the reanalysis performances.

This statistical evaluation serves as a gauge to measure the benefits expected from future developments of our current framework, and by that way the future evolutions of our long range module and methodology of use in case of an emergency response, and helps to determine the relevance of alternative simulation options regarding key parameters or processes, such as exhalation and soil moisture. A well-validated framework is of interest as to assess outdoor concentrations of Radon-222.

How to cite: Quérel, A., Meddouni, K., Quélo, D., Doursout, T., and Chuzel, S.: Statistical approach to improve Radon-222 long-range atmospheric transport modelling, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-9081, https://doi.org/10.5194/egusphere-egu21-9081, 2021.

Displays

Display file