Mitigation of urban storm water runoff through application of computational fluid dynamics

Doktorsavhandling, 2021

This thesis covers computational methods for improving the ability of green roofs to mitigate storm water runoff in urban environments. Roofs with living vegetation, known as green roofs, have been used for this purpose however quantification of their ability to slow and stop rainfall runoff has not been undertaken to a large degree. In this work two different approaches are taken: i) to improve green roof performance by optimizing their location on a building facade; and ii) to optimize the design of the growth substrate by examining the impact of the porous microstructure on infiltrating flow. The approach for optimization by placement makes use of traditional computational fluid dynamics and applies a full turbulence model to an Eulerian multiphase system consisting of a steady-state wind phase and a set of transient rainfall phases. The rainfall phases are determined by droplet size and the quantity incident upon the building facade is calculated and compared to experimental results. The analysis shows that the accuracy varies widely dependent upon location upon the structure and several sources of error are discussed. The second approach makes use of the lattice Boltzmann technique to aid in the deisgn of the growth substrate. Several representative porous media are generated using monodisperse randomly packed particles and gravity-driven infiltration is tracked from an initialized standing water height above the porous subdomain. Many aspects of the flow and properties of the microstructure are analyzed and conclusions are drawn based upon such factors as interfacial area, saturation rate, capillary pressure, and pore size distribution. Guidelines are drawn to aid in the design of thin homogeneous growth substrates based upon the findings. These ideal cases are compared to simulations performed on XMT scans of real growth substrate material and some conclusions are drawn on the observed differences.