Atmospheric dispersion of reactor-released 41Ar on the near range: A model intercomparison study

Accurate prediction of near-range atmospheric dispersion is essential for assessing the impact of small modular reactors, yet most operational models are not designed for short distances. This study compares four models – ADDER, FLEXPART, RIMPUFF, and URD – using 41 Ar release data from the Belgian Reactor 1 in Mol. The signal from this 41 Ar plume was recorded by a network of on-site environmental radiation monitors during a controlled release. The models were benchmarked against the monitoring network. After careful tuning of the models for near-range application, the models produced consistent results for time-integrated concentrations and ambient dose estimates. Key differences arose from meteorological inputs: simulations driven by on-site data better captured local variability, whereas NWP-driven FLEXPART could not resolve fine-scale wind structures. Maximum time-integrated concentrations (TICs) varied by 51% between the lowest (RIMPUFF) and the highest predictions (ADDER) while maximum doses differed by 39% between the lowest (FLEXPART) and the highest predictions (URD). Comparison with observations confirmed reasonable agreement, while also highlighting model-specific limitations in reproducing extremes. These findings underscore the importance of careful adaptation and benchmarking of operational models for reliable near-range emergency preparedness. • Operational models need parameter tuning to correctly model dispersion at near range. • After parameter tuning, all models produce comparable time-integrated doses. • Model performance varies when predicting (the extremes in) dose rate time traces. • Inter-model differences are driven by dispersion schemes and meteorological inputs.