CFD dataset for CityTransformer

This dataset contains the simulated plume concentration data in the urban area for CityTransformer, where all the data are created by 3D Large Eddy Simulations dedicated to the Oklahoma city. A single netcdf file in the dataset consists of the 2D image-like data on the ground (e.g building shapes, source locations, plume concentration) and 1D time-series data (e.g flow velocities, plume concentration) from a few monitoring stations. CityTransformer is designed to give the emergence contaminant dispersion prediction in a realistic urban area based on the deep learning model.

We simulate a 4 km x 4 km x 2.5 km area of Oklahoma city under the uniform flow conditions. In each simulation, 25 tracer particles are released from random locations in the city center. We consider three distinct 1.5 hour long time windows for data augmentation, wherein we sampled 1-hour long time varying 1D data and time-averaged 2D ground data for the next 30-minutes. The details of the dataset is described in the Github page.