Impact of modeling tool choice on ERPG safety zones for ammonia releases

Accurate prediction of the consequences of accidental releases of hazardous substances is essential for defining safety distances and emergency planning zones. Several software tools formally recognized by regulatory frameworks are available for use in safety studies. However, they often rely on different theoretical formulations and simplifying assumptions, which may lead to discrepancies in the predicted hazard footprints for identical release scenarios. This study compares toxic dispersion predictions from PHAST v9.0 and EFFECTS v12.5 for representative release scenarios under 1.5F and 5D meteorological conditions. The analysis focused on discrepancies between Emergency Response Planning Guidelines (ERPGs) concentration footprints at 1 m above the ground level, evaluated using similarity and distance metrics, including Jaccard and Dice indices, as well as Hausdorff and Chamfer distances. Results showed substantial differences under default software configurations in both source-term representation and predicted footprints, with Jaccard indices ranging from 0.35 to 0.69 and maximum downwind discrepancies exceeding 1.7 km in some cases. Harmonizing the representative release rate reduced part of this discrepancy, increasing footprint similarity by up to 10 % for the 5D condition. The findings emphasize the need to raise awareness among practitioners and regulators regarding variability arising from tool choice and underlying modeling assumptions. Quantifying these differences supports transparency, consistency in safety studies, accurate safety zone definition, and the harmonization of risk assessment practices.

Keywords: Toxic gas dispersion, Integral models, PHAST, EFFECTS, Similarity metrics