Pressure-liquefied ammonia jet dispersion: Multi-model intercomparison using Desert Tortoise and FLADIS field data

This paper presents the findings of an international model inter-comparison exercise that was undertaken in the period 2021-2024 to assess the performance of atmospheric dispersion models for simulating releases of pressure-liquefied ammonia. The exercise used data from ammonia field trials dating from the 1980s and 1990s: the Desert Tortoise and the FLADIS trials. Concentration data from two arcs of sensors in the Desert Tortoise trials and three arcs of sensors in the FLADIS trials were used. Twenty-one independent modelling teams from North America and Europe participated in the exercise and provided in total twenty-seven sets of results from a range of different models, including empirically-based nomograms, integral, Gaussian puff, Lagrangian particle, and Computational Fluid Dynamics (CFD) models. The work is novel in presenting the results from such a large cohort of models, examining specifically the dispersion behaviour of ammonia. This is particularly relevant at the current time, given the growing international interest in using ammonia as a clean energy vector and shipping fuel. The study found that the agreement between model predictions and measurements (as determined by performance measures such as geometric mean bias and geometric variance) varied between different models. At any downwind distance, the range in predicted plume arc-max concentrations spanned a range of up to one or two orders of magnitude about the measurements. Several modelling teams used the same models and, in most cases, their predictions differed. Given appropriate inputs, most models generally predicted concentrations that agreed with the data within commonly-used model acceptance criteria. There was no single class of model that provided superior predictions to others; predictions from several empirically-based nomograms, integral, Gaussian puff, Lagrangian particle, and CFD models were all in close agreement with the data (as defined by the model acceptance criteria). The findings of the exercise are being used to help plan a programme of future ammonia experiments in the USA, called the Jack Rabbit III trials. The results are also useful for assessing the performance of models that may be applied to assess risks at ammonia facilities, and for emergency planning and response. • Study assessed capabilities of dispersion models for simulating ammonia releases • Model intercomparison exercise used data from the Desert Tortoise and FLADIS trials • 21 participating organisations and 27 sets of results from range of different models • Most models predicted acceptable arc-max concentrations • Findings useful for risk assessments, emergency response and planning future trials