Uncertainty quantification in long range lagrangian atmospheric transport and dispersion modelling

De Meutter Pieter


Termonia Piet, (UGent), Piet.Termonia@UGent.be

SCK•CEN Mentor

Camps Johan
+32 14 33 27 61

Expert group

Crisis Management and Decision Support

PhD started



Measurements from networks monitoring for abnormal levels of airborne radioactivity such as the radionuclide network of the International Monitoring System of the preparatory commission of the Comprehensive Nuclear Test Ban Treaty Organization (CTBTO) to detect nuclear explosions, often need to be supplemented with the information on the regions where the detected radioactivity will be transported in the future or where it was present in the past. Such information can be provided by simulations performed with an atmospheric dispersion model. A dispersion model, however, being a mathematical and numerical representation of physical phenomena is subject to approximations made at all levels of its construction. These approximations necessarily introduce uncertainties, which influence quality of model outputs. The main objective of the proposed Ph.D. topic is to provide a numerical uncertainty quantification finally able to be translated into an error bar attached to the outputs of an atmospheric dispersion model.

Such an uncertainty quantification is also indispensable in case of a nuclear emergency so that a reliable upper bound can be derived for the forecast peaks of activity concentration or the dose received by the population. Likewise, it is essential for backtracking calculations as it helps establishing a robust borderline of the regions where the source of the detected radionuclides could have been located. The challenge resides in complexity of the underlying physical processes, which implies a convoluted interplay of model inputs and parameterizations. An important source of uncertainty is the uncertainty of changing weather conditions and/or the numerical weather prediction data. However, a framework spanned by a known emission inventory and trustworthy measurements as available from the International Monitoring System is an invaluable asset as it offers a rare possibility of disentangling uncertainties in the atmospheric dispersion model from other factors.


In collaboration with the Royal Meteorological Institute and the CTBTO the dispersion model FLEXPART with Numerical Weather Prediction input from the ECMWF at a global scale and a horizontal resolution of 0.5 degrees will be further made operational to perform the necessary calculations for the uncertainty quantification.

A study will be made on the sensitivity of the model results as a function of model parameters over the realistic physical range. In this way most important factors influencing the model uncertainty will be identified. 

In order to assess the uncertainty of changing weather conditions on the modelling dispersion results, FLEXPART will to be set-up to run with the 50 members of the ECMWF Ensemble Prediction System (EPS), which are intelligently designed to take into account uncertainty in numerical weather forecast. This will make it possible to assess for each species under consideration an uncertainty analysis.

To use this uncertainty information in an operational context, an algorithm will be designed to incorporate this NWP uncertainty for the species under consideration in the operational model which only uses the deterministic ECMWF input, since it is computationally very expensive to run the dispersion model ingesting the 50 EPS members into FLEXPART.

To contribute to the identification of the most important sources of uncertainty and to determine in the end the relevance of the uncertainty quantification one or more cases will be analyzed. These cases can include the North-Korean nuclear bomb tests, the Fukushima nuclear accident and releases from radiopharmaceutical facilities in normal operation. For all of these cases high quality data from the radionuclide component (including noble gas component) of the International Monitoring System can be used.