| Description |
QUIC EnvSim (QES) is a complete building-resolving urban microclimate modeling system developed to rapidly compute mass, momentum, and heat transport for the design of sustainable cities. One of the more computationally intensive components of this type of modeling system is the transport and dispersion of scalars. In this paper, we describe and evaluate QESTransport, a Reynolds-averaged Navier-Stokes (RANS) scalar transport model. QESTransport makes use of light-weight methods and modeling techniques. It is parallelized for Graphics Processing Units (GPUs), utilizing NVIDIA's OptiX application programming interfaces (APIs). QESTransport is coupled with the well-validated QUIC Dispersion Modeling system. To couple the models, a new methodology was implemented to efficiently prescribe surface flux boundary conditions on both vertical walls and flat surfaces. In addition, a new internal boundary layer parameterization was introduced into QUIC to enable the representation of momentum advection across changing surface conditions. QESTransport is validated against the following three experimental test cases designed to evaluate the model's performance under idealized conditions: (i) flow over a step change in moisture, roughness, and temperature, (ii) flow over an isolated heated building, and (iii) flow through an array of heated buildings. For all three cases, the model is compared against published simulation results. QESTransport produces velocity, temperature, and moisture fields that are comparable to much more complex numerical models for each case. The code execution time performance is evaluated and demonstrates linear scaling on a single GPU for problem sizes up to 4.5 x 4.5 km at 5 m grid resolution, and is found to produce results at much better than real time for a 1.2 x 1.2 km section of downtown Salt Lake City, Utah. |
