Identification and verification of worst-case radiological transport scenarios for Ireland: a simulation-based approach to nuclear emergency preparedness (2011-2024)

Marc Sturrock, Robert Ryan, Kevin Kelleher

This study presents a comprehensive simulation-based assessment of potential transboundary radiological transport to Ireland from six nuclear facilities in the United Kingdom and France, utilising weather data over a fourteen-year period (2011–2024). Systematic screening of 2.2 million HYSPLIT atmospheric dispersion simulations identified eighteen worst-case scenarios representing conditions of maximum ground deposition, maximum air concentration, and minimum warning time for protective action implementation. Independent verification using FLEXPART and HYSPLIT demonstrated expected inter-model variability (factor of 1–10), with both Lagrangian models providing consistent risk assessment brackets. Heysham, despite its complex 19-isotope AGR source term, produced negligible radiological doses to Ireland (< 0.01 mSv)–more than four orders of magnitude below intervention thresholds. More distant continental facilities (Flamanville, Paluel, Sizewell B) showed low but measurable doses (0.1–4.6 mSv depending on scenario and model), remaining well below the 50 mSv sheltering threshold. Hinkley Point C (under construction) showed elevated but sub-threshold doses (0.3–8.5 mSv depending on model). However, the cancelled Wylfa Newydd gigawatt-scale project (the site is now proposed for small modular reactors), owing to its extreme proximity to Ireland, exhibited concerning dose predictions: FLEXPART calculated 20.7 mSv under maximum deposition conditions (May 2024 scenario), approaching the 50 mSv sheltering threshold, whilst HYSPLIT predicted 4.5 mSv. This inter-model variability (factor of ~5) highlights genuine uncertainty for near-source impacts but converges on a critical finding: were a gigawatt-scale reactor constructed at the Wylfa site, severe accidents during specific meteorological patterns could require protective actions in Ireland. Machine learning models (XGBoost) achieved validation accuracies of 85–93% for rapid impact prediction, whilst global sensitivity analysis revealed that meteorological conditions, rather than release parameters, dominate consequence severity. These findings provide quantitative assurance that existing nuclear infrastructure poses low transboundary risk to Ireland well below intervention thresholds, whilst demonstrating that facility proximity constitutes the dominant factor determining potential radiological impact.