The increase in computational resources has allowed to couple mesoscale numerical weather prediction models (WRF), having typical grid resolutions of kilometric size, with microscale mass-consistent or Computational Fluid Dynamics (CFD) codes solving for wind over complex terrains and/or at building scales. Urban-scale wind and atmospheric dispersal simulations require of accurate geometric descriptions of buildings and underlying terrain. For this, we combine Digital Elevation Models (DEM) for the underlying terrain, a cadastre of the target city (or metropolitan area) to limit the city blocks/buildings and LiDAR data to retrieve building elevation above terrain. The automatic procedure allows for a high-resolution computational mesh of a city used to solve for hourly winds and pollutant dispersion.