A sensitivity analysis applied to SPRAY and CALPUFF models when simulating dispersion from industrial fires

This paper discusses a hypothetical case study in which SPRAY and CALPUFF dispersion models are applied to the simulation of an incidental fire. For this type of accident, source features are typically not directly measurable, thus making their definition critical. The choice of some model-specific parameters is another critical issue, since clear indications are rarely available in guidelines. The aim of this work is to compare how pollutant concentrations simulated with the two models are affected by changing these two sets of data (i.e. parameters related to the emission source and model specific parameters), thus performing a sensitivity study to identify the most influential variables. The most relevant outcome is that sensitivities of the two models are generally comparable, except for the source diameter: if the SPRAY model is applied with the specific fire source option, then the concentrations result almost independent from this parameter. Conversely, when considering other source-types, the concentrations vary up to ± 60% within the selected uncertainty range.