GenDA: Generative Data Assimilation on Complex Urban Areas via Classifier-Free Diffusion Guidance

2026-01-16

AI Summary

Accurate reconstruction of high-resolution urban wind fields from sparse sensor observations remains highly challenging, limiting reliable assessment of air quality and thermal environments. This work proposes GenDA, a framework based on a multi-scale graph diffusion model trained on CFD data, which efficiently reconstructs wind fields on unstructured grids by integrating geometric priors and observational constraints through a classifier-free guidance mechanism. The approach interprets classifier-free guidance as a posterior reconstruction process, enabling zero-shot generalization to unseen building layouts, wind directions, and grid resolutions without retraining, while uniformly handling both fixed-point and trajectory-based observations. Experiments demonstrate that GenDA reduces relative root-mean-square error by 25–57% and improves structural similarity index by 23–33% compared to GNN-based supervised models and traditional reduced-order data assimilation methods, with validation on real-world urban RANS simulations.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Web data generation and simulation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Urban wind flow reconstruction is essential for assessing air quality, heat dispersion, and pedestrian comfort, yet remains challenging when only sparse sensor data are available. We propose GenDA, a generative data assimilation framework that reconstructs high-resolution wind fields on unstructured meshes from limited observations. The model employs a multiscale graph-based diffusion architecture trained on computational fluid dynamics (CFD) simulations and interprets classifier-free guidance as a learned posterior reconstruction mechanism: the unconditional branch learns a geometry-aware flow prior, while the sensor-conditioned branch injects observational constraints during sampling. This formulation enables obstacle-aware reconstruction and generalization across unseen geometries, wind directions, and mesh resolutions without retraining. We consider both sparse fixed sensors and trajectory-based observations using the same reconstruction procedure. When evaluated against supervised graph neural network (GNN) baselines and classical reduced-order data assimilation methods, GenDA reduces the relative root-mean-square error (RRMSE) by 25-57% and increases the structural similarity index (SSIM) by 23-33% across the tested meshes. Experiments are conducted on Reynolds-averaged Navier-Stokes (RANS) simulations of a real urban neighbourhood in Bristol, United Kingdom, at a characteristic Reynolds number of $\mathrm{Re}\approx2\times10^{7}$, featuring complex building geometry and irregular terrain. The proposed framework provides a scalable path toward generative, geometry-aware data assimilation for environmental monitoring in complex domains.

Problem

• urban wind flow reconstruction
• sparse sensor data
• data assimilation
• complex urban geometry
• high-resolution wind fields

Innovation

• Generative Data Assimilation
• Classifier-Free Diffusion Guidance
• Graph-Based Diffusion Model
• Unstructured Mesh Reconstruction
• Urban Wind Flow