Evaluation of a neighbourhood scale, street network dispersion model through comparison with wind tunnel data

Abstract

This study compared dispersion calculations using a street network model (SIRANE) with results from wind tunnel experiments in order to examine model performance in simulating short-range pollutant dispersion in urban areas. The comparison was performed using a range of methodologies, from simple graphical comparisons (e.g. scatter plots) to more advanced statistical analyses. A preliminary analysis focussed on the sensitivity of the model to source position, receptor location, wind direction, plume spread parameterisation and site aerodynamic parameters. Sensitivity to wind direction was shown to be by far the most significant. A more systematic approach was then adopted, analysing the behaviour of the model in response to three elements: wind direction, source position and small changes in geometry. These are three very critical aspects of fine scale urban dispersion modelling. The overall model performance, measured using the Chang and Hanna (2004) criteria can be considered as ‘good’. Detailed analysis of the results showed that ground level sources were better represented by the model than roof level sources. Performance was generally ‘good’ for wind directions that were very approximately diagonal to the street axes, while cases with wind directions almost parallel (within 20°) to street axes gave results with larger uncertainties (failing to meet the quality targets). The methodology used in this evaluation exercise, relying on systematic wind tunnel studies on a scaled model of a real neighbourhood, proved very useful for assessing strengths and weaknesses of the SIRANE model, complementing previous validation studies performed with either on-site measurements or wind tunnel measurements over idealised urban geometries.