Comparison of atmospheric dispersion simulations in the near field (<200) with operational Lagrange model and CFD methods

This study evaluates two models for simulating near-field (<200 m) atmospheric dispersion: an operational Lagrangian model, and Computational Fluid Dynamics (CFD) simulations, including Reynolds-Averaged Navier-Stokes (RANS) and Large Eddy Simulation (LES) approaches. The evaluation is conducted against data from a full-scale atmospheric tracer experiment.

The first  model is the Safety Lagrangian Atmospheric Model (SLAM)[1], which implements a Lagrangian stochastic particle dispersion framework coupled with pre-calculated wind and turbulence fields derived from a series of RANS simulations performed with ANSYS Fluent CFD code. On the other hand, the CFD dispersion simulations employ the open-source CALIF³S solver including both RANS and LES methodologies[2] The turbulence closure models correspond respectively to a  standard two-equation RANS model and a hybrid RANS/LES approach based on a Detached-Eddy-Simulation (DES) methodology. The full-scale atmospheric dispersion experiment DIFLU (Dispersion du Fluor 18 en Milieu Urbain)[3], provides the validation dataset. The experiment includes tracer concentration measurements under various meteorological conditions within 500 meters of a cyclotron facility.

The results demonstrate that SLAM predicts concentration values comparable to those obtained by CALIF³S-RANS despite the distinct inlet boundary conditions. SLAM uses similarity theory to calculate the inlet velocity, temperature and turbulent profiles, whereas inlet profiles used in CALIF³S are extracted from precursor simulations that best match the measurements. In the near-source region (<50 m), where turbulence plays a significant role, both CALIF³S-RANS and SLAM overestimate the concentration distributions compared to those given by CALIF³S- LES approach. LES results are closer to the measurement because of its advantage in predicting eddy detachments. The disparity of concentration values between RANS and LES diminishes rapidly beyond 100 m from the source, where the region is free of buildings. Overall, SLAM and CALIF³S-LES achieve similar performance in terms of the fraction of simulated values within a factor of two of the measurements under neutral atmospheric conditions. Hence, using RANS simulation is sufficient to achieve acceptable results in the near-field dispersion simulation (<200m) under neutral atmospheric conditions. Nevertheless, more precise results can be achieved with LES method in the near source region (<50m). Future work will focus on other experimental cases, including unstable atmospheric conditions.

 

[1]        S. Yang, I. Korsakissok, P. Laguionie, and P. Volta, “Dispersion du FLuor 18 en milieu Urbain near-field (>200m) atmospheric dispersion simulation and sensitivity analysis following a full scale atmospheric tracer experiment,” in 22nd Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes

[2]        J. Janin, F. Duval, C. Friess, and P. Sagaut, “A new linear forcing method for isotropic turbulence with controlled integral length scale,” Phys. Fluids, vol. 33, no. 4, Apr. 2021, doi: 10.1063/5.0045818/1065646.

[3]        P. Laguionie et al., “Investigation of a Gaussian Plume in the Vicinity of an Urban Cyclotron Using Helium as a Tracer Gas,” Atmosphere (Basel)., vol. 13, no. 8, pp. 1–16, 2022, doi: 10.3390/atmos13081223.