This thesis studied urban particle dispersion close to source and ground. It used an existing, steady state, three dimensional Lagrangian particle dispersion model, which includes roughness sublayer parameterizations of turbulence and flow. The model is valid for convective and neutral to stable conditions and uses the kernel method for concentration calculation.
Already published corrections to the original model formulation were introduced to the model. An additional model error in the velocity auto-covariance derivative parameterization of the roughness sublayer, detected during the course of this work, was corrected. The impact of these changes were compared to that of a new dissipation rate parameterization in the roughness sublayer, which was based on observations. Furthermore, an earlier work hypothesized that improving the lower boundary condition could improve the model predictions. This modification was realized and its influence compared to the other changes. To reach these goals, the model was initialized and compared with measurements that had been taken during the Basel UrBan Boundary Layer Experiment (BUBBLE), using SF6 as tracer.
One of the corrections to the model formulation, changing the solenoidal probability current, provides better model results, while the rest were only wrong in the publications, not the source code. The newly detected model error has a large influence on the model results. Model performance was neither clearly enhanced nor was it definitely diminished by the new dissipation rate parameterization. Despite this, the present results indicate that the magnitude of the dissipation rate is more important than the shape and derivative of its profile in the roughness sublayer. After correcting the newly detected model error, the results that inspired the modification of the lower boundary condition could no longer be replicated. Consequently, the proposed modification does not improve the model’s agreement with measurements. Additionally, this thesis showed that the model with the drift boundary condition is similarly sensitive to adjustments of the zero plane displacement height and the roughness sublayer boundary height as the original model.