Comparative metrics for computational approaches in non-uniform street-canyon flows

摘要

Three different computational fluid dynamics (CFD) methods are assessed for their ability to predict topological flow features in idealized street canyons with uneven building heights. Mean velocity-fields from step-up (i.e., a high-rise building downwind of a low-rise building) and step-down (i.e., a low-rise building downwind of a high-rise building) street canyons are evaluated against high-spatial-resolution wind-tunnel data. Each method represents a different level of flow physics using: a mass-conserving model entitled Quick Urban & Industrial Complex wind model (QUIC-URB), a Reynolds-averaged Navier-Stokes (RANS) model, and a large-eddy simulation (LES) model. A new metric that represents the equally weighted trade-off between accuracy and efficiency is introduced to evaluate the CFD methods' capabilities to capture major-flow topological features in uneven building height street canyons. For step-up street canyons, all three methods qualitatively predict primary topological features, however, none simultaneously capture all secondary features. For step-up street canyons and step-down street canyons with narrow-streets, QUIC-URB captures most of the primary flow topological features including stagnation and saddle points and rooftop recirculation zones. RANS captures primary vortices for step-up street canyons and step-down street canyons with wide-streets. LES is computationally costly but it is the only method that successfully predicts secondary flow topological characteristics for step-down street canyons with wide-streets. When examining our trade-off metric, QUIC-URB has the highest score for step-up street canyons, while QUIC-URB and RANS have equally high trade-off scores for step-down street canyons with narrow-streets, and QUIC-URB and LES have nearly equal trade-off scores for step-down street canyons with wide-streets.