Multiscale simulation of the urban wind environment under typhoon weather conditions

Accurate simulations of planetary boundary layer (PBL) winds in urban areas require combining meteorological knowledge and fine-grained geometrical information. Computational fluid dynamics (CFD) is widely used to assess pedestrian wind comfort and wind disasters in planning resilient cities. However, the CFD-predicted PBL is highly affected by the inflow boundaries. Wind profiles under extreme weather conditions, such as tropical cyclones, can hardly be determined, and associated uniform logarithmic or power law expressions have not been obtained. In this study, urban wind flow over mountainous terrain was simulated using a one-way nested simulation approach between mesoscale and microscale models. The inflow wind speed, turbulence scalars, and potential temperature in the CFD code are sustained by the numerical weather prediction (NWP) model. Methodologies considering typhoon weather conditions were examined to improve the numerical accuracy in determining mesoscale typhoon structures and pedestrian-level wind conditions. The numerical errors were quantified in mesoscale and microscale formulations. A new tendency assimilation was proposed by incorporating local-scale observations into the CFD domain. This approach entailed empirical mode decomposition to quantify the mean wind speed differences between the observations and NWP results, which were then extrapolated to NWP-CFD nested interfaces via multiplication by the spatial correlation coefficient. The numerical performance was validated against both on-site observations for meteorological purposes and wind profiles retrieved from the experimental LiDAR of the landfalling typhoon Haima. The simulated wind field exhibited an increased accuracy in the local urban area. More specifically, the index of agreement in wind speeds was improved from 0.28 to 0.72, and the mean absolute errors were reduced from 5.46 m/s to 1.89 m/s.