The interaction between the surface and atmosphere within Urban Boundary Layers (UBLs) at the local scale, particularly in densely built environments, is inherently complex due to the heterogeneity of surface roughness. This interaction plays a crucial role in shaping urban weather phenomena, defining micro-climates, and influencing energy consumption, thermal comfort, and the dispersion of aerosols and air pollution. To understand this complexity, a high-resolution, computationally efficient, and reliable UBL mixing Computational Fluid Dynamics (CFD) model is essential. This study introduces a surrogate Reduced Order Model (ROM) that addresses the computational challenges of CFD for UBL and demonstrates its broad applicability. Using a representative urban test case from central Paris, we illustrate the effectiveness of our approach. In the offline phase, the ROM is constructed by assembling a database of Streaming Dynamic Mode Decomposition (sDMD) modes related to various aspects of UBL dynamics, such as temperature distribution, wind patterns, and turbulence characteristics. In the online phase, we interpolate these DMD atmospheric regimes from the database, allowing us to determine the dynamic characteristics of the UBL within the domain without initiating computationally intensive CFD calculations. The results demonstrate that the ROM can accurately predict UBL dynamics across a range of meteorological conditions, with an RMSE of less than 8\% for key variables. Additionally, the ROM achieves a significant reduction in computational time, with a speedup of up to 1000 times compared to traditional CFD simulations. This study highlights the potential of ROMs as versatile tools for real-time simulations and large-scale applications in urban meteorology and air quality modeling.