Catalog Home Page

Urban dispersion modelling in an emergency response context

Donnelly, R.P. (2010) Urban dispersion modelling in an emergency response context. PhD thesis, Murdoch University.

[img]
PDF - Whole Thesis
Available Upon Request

Abstract

Chemical and biological releases into the atmosphere induce impacts over several spatial scales, both close to the source and potentially many kilometres downwind. Whilst regional scale impacts are well documented and modelled, pollutant dispersion within the immediate local scale, particularly in urban areas, is not well represented. Yet this is the region close to the source where the most severe effects from chemical and biological releases can occur.

Both from an emergency response perspective in maintaining order, defining safe-zones and performing rescue and evacuation, as well as for long term community health, it is essential to have a well based understanding of pollutant dispersion in these complex urban environments and the ability to respond quickly to accidental spills. Thus, the objective of this study is to develop an understanding of the building-to-regional scale dispersion of pollutant within the Perth central business district (CBD) through direct tracer experiments in conjunction with a suite of computational models to define and refine appropriate limits of currently used emergency response models.

A series of releases of sulphur hexafluoride (SF6) were undertaken within the Perth CBD and sampled in Tedlar bags over an extended grid before undergoing analysis by gas chromatography with electron capture detection (GC/ECD). This provided near field concentration distributions over selected periods under different stability regimes and seasons. Combined with the Mock Urban Setting Test (MUST) data, these provided idealised and real urban data sets for model comparison and validation. Three models were assessed against these data; a standard Gaussian emergency response model, a Lagrangian puff model and the computational fluid dynamics package WinMISKAM.

Under the idealised circumstances of MUST, comparisons were made between observed ground level concentrations and predictions generated by WinMISKAM. The model was driven by 5 min averaged on-site meteorological data, and used minimum grid spacing of 0.5 m in both the horizontal and vertical. It was found to perform well, with 46% of all predictions (paired in time and space) and 83% of arc maxima predictions within a factor of two of observed concentrations. The model performed better for neutral cases than stable cases with 27% of stable case predictions and 57% of neutral case predictions within a factor of two when compared in time and space. Whilst the meteorological pre-processor in the Lagrangian puff model reproduced the observed MUST friction velocity and Monin-Obukhov lengths, the model predicted broader plumes than are seen in the observations, resulting in underprediction of peak concentrations and a nett overprediction of concentrations by approximately a factor of two against all ground level data.

Against the CBD dataset, two Gaussian techniques, similar to current emergency response approaches, were used and these poorly reflected trends in the observations, with 8.9% and 10% of predictions within a factor of two respectively and false negative predictions of 33% and 19% of all data points, respectively. Winiii MISKAM model comparison showed false negatives for 19% of all data points and 17% of predictions within a factor of two. The narrower modelled plumes generated by WinMISKAM resulted in it performing better when the mean, above canopy wind is along-canyon, rather than across-canyon and underpredicting observed concentrations by approximately a factor of 1.8 for all data. Sixteen percent of the the Lagrangian puff predictions were within a factor of two of Perth CBD dataset observations. This model has been shown to be more suited to emergency response applications than the other models investigated, not only due to the significantly lower run-times required by this model, but also due to the nett overprediction of approximately a factor of 1.4. This, along with a much lower propensity toward false negative predictions (11%) results in more conservative predictions being provided to emergency responders. Thus a simple Lagrangian model has been shown to provide superior predictions than those currently available in emergency response situations, i.e. from a Gaussian model, and to approach the level of accuracy obtained using more sophisticated modelling techniques, within a timeframe useful to emergency responders.

Item Type: Thesis (PhD)
Murdoch Affiliation: School of Environmental Science
Notes: Note to the author: If you would like to make your thesis openly available on Murdoch University Library's Research Repository, please contact: repository@murdoch.edu.au. Thank you.
Supervisor(s): Lyons, Tom
URI: http://researchrepository.murdoch.edu.au/id/eprint/41610
Item Control Page Item Control Page