Dispersion Simulation and Visualization for Urban Security Feng Qiu Ye Zhao Zhe Fan Xiaoming Wei Haik Lorenz Jianning Wang Suzanne Yoakum-Stover Arie Kaufman Klaus Mueller ∗ Center for Visual Computing and Department of Computer Science Stony Brook University, Stony Brook, NY 11794-4400 ABSTRACT We present a system for simulating and visualizing the propaga- tion of dispersive contaminants with an application to urban secu- rity. In particular, we simulate airborne contaminant propagation in open environments characterized by sky-scrapers and deep urban canyons. Our approach is based on the Multiple Relaxation Time Lattice Boltzmann Model (MRTLBM), which can efficiently han- dle complex boundary conditions such as buildings. In addition, we model thermal effects on the flow field using the hybrid thermal MRTLBM. Our approach can also accommodate readings from var- ious sensors distributed in the environment and adapt the simulation accordingly. We accelerate the computation and efficiently render many buildings with small textures on the GPU. We render stream- lines and the contaminant smoke with self-shadowing composited with the textured buildings. Keywords: Lattice Boltzmann Model, GPU, Visualization 1 I NTRODUCTION According to the National Research Council report Tracking and Predicting the Atmospheric Dispersion of Hazardous Releases, “Our nation’s capacity to respond to atmospheric C/B/N (chemi- cal/biological/nuclear) events stands, like a three-legged stool, on the strength of three interconnecting elements: (1) dispersion mod- els that predict the path and spread of the hazardous agent; (2) ob- servations of the hazardous plume itself and of local meteorologi- cal conditions; and (3) interaction with emergency responders who use the information provided by the models.” The simulation work we present in this paper is directly relevant to the first and third elements. The Lattice Boltzmann Model (LBM) that we use can accurately model air flow and contaminant transport and mixing in geometrically complex environments with the inclusion of thermal effects due to surface heating. By exploiting the inherent locality of the LBM and implementing the computation on the GPU, we further demonstrate that it is feasible to build large scale simula- tions that span a whole city. We also show the performance and visualization advantages that result from using the GPU for scien- tific computation. The importance of visualization stems from its ability to enhance the usefulness and accessibility of the informa- tion provided by the model. Our demonstration application illus- trates how the combination of LBM modeling and GPU compu- tation can enhance our understanding of meteorological and fluid dynamic processes governing dispersion in urban areas and also al- low emergency management, law enforcement and other personnel to adequately plan for, train for, and respond to potential accidents or attacks involving toxic airborne contaminants. ∗ Email:{qfeng, yezhao, fzhe, wxiaomin, hlorenz, jianning, suzi, ari, mueller}@cs.sunysb.edu 2 RELATED WORK Researchers have conducted dispersion observation experiments in various environments. The urban tracer and meteorological field campaign (URBAN) conducted in Salt Lake City in 2000 investi- gated meteorological and fluid dynamic processes governing dis- persion in urban environments. In particular, the study attempted to resolve interacting scales of atmospheric motion from the scale of individual buildings to that of whole cities and entire regions [2]. Another meteorological field campaign conducted during Oc- tober 2000 in the Salt Lake Valley studied vertical transport and mixing (VTMX) processes [13]. The focus of that project was to measure, characterize and analyze VTMX processes, especially in urban areas larger than that of URBAN 2000. The data and insights resulting from these campaigns will help to build better models and evalute the performance of existing numerical simulations for dis- persion in urban environments. In terms of modeling, Pardyjak et al. [31, 30], Williams et al. [42, 43] and Boswell et al. [6] have proposed a fast-response ur- ban dispersion modeling system that computes 3D wind patterns and dispersion of airborne contaminants in urban areas with many buildings. The wind model (QUIC-URB) uses empirical algorithms that estimate the wind fields around buildings. The Lagrangian dispersion model (QUIC-PLUME) computes the dispersion using random walk equations based on the mean wind field produced by QUIC-URB. Brown et al. [7, 8] have presented a modeling ap- proach to compute wind fields and simulate the transport of agents in three different scales. A numerical weather prediction model called COAMPS [21, 10] computes the wind field and other me- terological physical effects such as temperature at the urban scale. At the many-building scale, HIGRAD [8, 34] computes the flow field around buildings and simulates contaminants transport. This model is a second-order accurate computational fluid dynamics (CFD) model based on the Navier-Stokes equations (NSE) with fi- nite difference approach. For single to few-building scale, another CFD model called FEM3MP [9] was used. This is a finite element model that can simulate a flow field and dispersion around indi- vidual buildings in great detail. The three models take appropriate scale-dependent physics into account and share data together. Recently, LBM [36] has been introduced to the graphics commu- nity for modeling various flow phenomena including wind, smoke, fire, and melting [38, 39, 41, 45]. Although LBM is a relatively new CFD procedure, it has the advantages of being simple to im- plement, parallelizable, and can accommodate complex boundaries. It can also be extended to model thermal effects, reactive flows, and other physics with relative ease. In contrast to the flow simula- tion methods described above, the LBM does not model the NSE directly. Rather, it models the micro-scale Boltzmann kinetics of fluid elements streaming and collision. As a numerical scheme, it is explicit, synchronous, second-order space-time accurate with an advection limited time-step. In the limit of zero time step and lattice spacing, LBM yields the NSE for an incompressible fluid. As a kind of explicit finite difference method, LBM is consistent for flows with low Mach number (i.e., flow velocities small compared to the