Answers are grounded in this paper and cite their sources.
Pollutant transport in complex urban environments is strongly shaped by building-induced flow distortion, channeling, recirculation, and geometrical blockage, which together produce highly heterogeneous concentration fields. Although high-fidelity numerical simulations can capture these effects, their computational cost limits rapid prediction and source inference. This study proposes a physics-informed surrogate modeling framework for rapid prediction and source inference of ground-level pollutant dispersion in complex urban environments under quasi-steady wind conditions. Four source-dependent flow-relative features are introduced to encode transport direction and geometry-constrained accessibility, namely along-wind distance, crosswind distance, line-of-sight indicator, and shortest-path distance. These features are used as inputs to deep-learning surrogate models, with DeepLabV3 + adopted as the primary architecture and compared against U-Net, FNO2d, and Pix2Pix. Experiments based on simulation data from eight realistic urban blocks in Paris and Grenoble show that the proposed feature encoding consistently improves cross-layout prediction, yielding R²(log) values of 0.72–0.85 across the tested architectures; eight-fold leave-one-block-out validation gives an average R²(log) of 0.80. When coupled with a response-library-based Bayesian inference procedure, the framework achieves an approximately 96% finding rate, compared with about 46% for a Gaussian-plume baseline. Response-library construction for a new block requires less than 1 min, while each online inference step takes about 1 ms. Additional validation against ANSYS Fluent cases gives high R²(log) values while maintaining successful source finding in both cases. The results indicate that physics-informed surrogate modeling provides an efficient route for rapid prediction and source inference of urban pollutant dispersion in complex built environments.
No takes yet. Share an insight, caveat, or question.
Z et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: