In this work, we employ a relatively new computational fluid dynamics model, the Lattice Boltzmann Method (LBM). Unlike other approaches, LBM discretizes the micro-physics of local interactions and can handle very complex boundary conditions, such as deep urban canyons, curved walls, indoors, and dynamic boundaries of moving objects. Furthermore, its computational pattern which is similar to Cellular Automata is easily parallelizable and hence can be accelerated on commodity graphics processing units (GPUs), achieving real-time or even accelerated real-time on ordinary PCs and laptops, providing a predictive tool for anticipating subsequent propagation. Another key innovation of LBM is its extension to support input from pervasive sensors. This will allow us to influence the simulation so as to maintain its faithfulness to real-time sensor readings.

We have implemented the parallel LBM computation on a cluster of GPUs. Our cluster, called the Stony Brook Visual Computing Cluster, has 32 nodes connected by a 1 Gigabit network switch. Each node is an HP PC equipped with a GPU, Nvidia GeForce FX 5800 Ultra. We have tested the LBM simulation with a GIS of Times Square Area of NYC, which consists of 91 blocks and roughly 850 buildings. We have also test it with a 10 block GIS around the EML building in the West Village of NYC, overlaid with results of dispersion simulation and real-time readings from 3 sensors installed on that building. In addition, we have implemented a 3D city navigation system (web-based or stand alone), featuring a 3D polygonal model GIS with façade texturing, flow visualization streamlines, volume rendering plumes, and information visualization of real-time sensor data.