The Third International Conference on Urban Air Quality

Abstract

Intercomparison of numerical urban dispersion models - Part I: street canyon and single building configurations
P. Sahm, P. Louka1, M. Ketzel2, E. Guilloteau3, J-F. Sini1

Laboratory of Heat Transfer and Environmental Engineering, Aristotle University, Thessaloniki, Greece
1Fluid Mechanics Laboratory, Ecole Centrale de Nantes, Nantes, France
2Department of Atmospheric Environment, National Environmental Research Institute, Roskilde, Denmark
3Institute of Hydromechanics, University of Karlsruhe, Karlsruhe, Germany


Microscale computational fluid dynamics (CFD) models explicitly taking into account the building structure have become a useful and popular tool for assessment and prediction of air quality in urban areas. The proper validation of such a model is a crucial prerequisite for its practical application. Within the framework of the European research network TRAPOS (see http://www.dmu.dk/AtmosphericEnvironment/trapos) a working group on computational fluid dynamics modelling was established and model intercomparison exercises were launched (http://www.dmu.dk/AtmosphericEnvironment/trapos/cfd-wg.htm). Different numerical models employing the widely used 'standard k-e -model' were applied to well defined test cases comprising a variety of 2 and 3 dimensional configurations for which measurements from wind tunnel or field studies were available.

This paper, which is the first in a sequence of two, is focused on the intercomparison of five models (CFX-TASCflow, CHENSI, CHENSI2, MIMO, MISKAM) in the cases of a single cavity and a surface mounted cube. The experimental database for the single-cavity case was established in the wind tunnel of the University of Surrey, whereas the cube experiment was carried out in the BLASIUS wind tunnel at the Meteorological Institute of Hamburg University.

The results of the model intercomparison indicate that the general agreement among the models is good with some differences occurring close to the solid boundaries. In order to determine the origin of the differences among the models results a detailed comparison between the model source codes is conducted focusing on the numerical scheme, boundary conditions and values of constants utilised. These two generally simple study cases are particularly interesting for the demonstration of the level of agreement expected between similar codes and for the deduction of useful recommendations for practical applications as well as for the improvement and further development of the models.