Toxic release dispersion modelling with PHAST : parametric sensitivity analysis

Abstract

Recent changes to French legislation, concerning the prevention of technological and natural risk, require industrial sites to calculate the safety perimeters for different accident scenarios, based on a detailed probabilistic risk assessment. It is important that the safety perimeters resulting from risk assessment studies are based on the best scientific knowledge available, and that the level of uncertainty is minimised. A significant contribution to the calculation of the safety perimeters comes from the modelling of atmospheric dispersion, particularly of the accidental release of toxic products. One of the most widely used tools for dispersion modelling in several European countries is PHASTTM[1]. This software application is quite flexible, allowing the user to alter values for a wide range of model parameters. Users of the software have found that simulation results may depend quite strongly on the values chosen for some of these parameters. While this flexibility is useful, it can lead different users to calculate effect distances that vary considerably even when studying the same scenario. In order better to understand the influence of these input parameters, we have carried out a parametric sensitivity study of the PHAST dispersion models. This allows us to obtain global sensitivity indices for the input parameters, which quantify the level of influence of each parameter on the output of the model and the interactions. The FAST (Fourier Amplitude Sensitivity Test) sensitivity analysis method that we have applied (using the SimLab software tool [2]) provides both first order indices (that characterize the parameter’s influence on the model output when it varies in isolation) and total indices (that characterize one parameter’s influence including its joint interaction with other input parameters). We shall present results of this analysis on a number of toxic gas dispersion scenarios. The analysis has considered parameters related to the physical release scenario (release rate, release height, etc.), to weather conditions (wind speed, stability class, atmospheric temperature, etc.) and to the numerical resolution (step size, etc.). We compare the results of several sensitivity analysis methods, both local one-at-a-time methods and global methods. We discuss the importance of selecting an appropriate model output value when studying the model sensitivity (output measures considered include the concentration of the released gas at a long distance, at a short distance, and the maximal distance at which a specified concentration is attained). Our experimental results assume that input parameters to the dispersion model are independent. However, correlations exist between several input parameters that we have analysed, such as wind speed and atmospheric stability class. We discuss various approaches to calculate sensitivity indices that take this correlation into account. [1] DNV Software, London, UK [2] Saltelli, A., Chan, K., Scott, E. M. Sensitivity Analysis, 2004, John Wiley & Sons Publishers.