The interaction between wind flow and complex urban morphology creates highly non-uniform velocity fields that are difficult to predict using standard meteorological data alone. We present a Computational Fluid Dynamics (CFD) analysis of urban wind flows to assess the vulnerability of street vegetation in southern Madrid. The study focuses on identifying critical wind zones where local acceleration puts vegetation stability at risk during extreme weather events. The computational domain was constructed using LiDAR and cadastral data to accurately capture the aerodynamic complexity of the Madrid cityscape, using a fully automated methodology. Steady-state simulations were performed using a RANS standard k-ε turbulence model, adhering to best practice guidelines for urban CFD simulations. Boundary conditions were defined by preserving the Atmospheric Boundary Layer equations at every point, by analyzing historical meteorological records to identify the two most statistically prevalent wind directions. These baseline scenarios were then contrasted with extreme velocity profiles reconstructed from recorded severe storm episodes to quantify aerodynamic amplification. This study compares velocity magnitude and turbulence intensity profiles between these scenarios. Results indicate that peak wind velocities are heavily concentrated in streets adjacent to open urban spaces and within street canyons aligned parallel to the prevailing wind vector, confirming significant channeling effects in these zones. By quantifying the amplification factors in these critical zones, this study provides a quantitative basis for assessing wind-throw risk in dense urban environments.