Simulation of smoke dispersion and air pollution dynamics

Abstract

A hybrid model approach is proposed to address the limitations of traditional physical and empirical models in simulating smoke diffusion and depicting gas pollution. This method integrates the semi-Lagrangian method for smoke modeling, with the k-d tree algorithm enhancing computational efficiency. To enhance smoke simulation details, a fluctuating wind field generated by the linear filter method is incorporated into the external force term to refine smoke particle trajectories. The rendering process employs a bidirectional projection function combined with actual smoke textures to mitigate the issue of particle visibility and to significantly enhance the visual fidelity of smoke diffusion. Additionally, an optimized Gaussian plume model is utilized to bridge the gap between physical and empirical models, while a pollution attenuation formula and refined Perlin noise are applied to enrich the global gas pollution details and to increase the realism of pollution dynamics. The method also incorporates an improved time axis algorithm to overcome the issue of static gas pollution color, enabling the depiction of dynamic and gradual pollution changes. Experimental results validate the effectiveness of this approach in generating realistic and dynamic gas pollution scenes in real-time.