TopoFlow: Physics-guided Neural Networks for high-resolution air quality prediction

2026-02-18

AI Summary

This study addresses the challenge of accurately modeling pollutant transport and dispersion under the influence of complex terrain and wind fields—a task where existing deep learning approaches often fail to adequately incorporate physical constraints. To this end, we propose TopoFlow, a vision Transformer architecture that explicitly embeds physical priors by integrating terrain-aware attention mechanisms and a wind-direction-guided patch reordering strategy. Trained on six years of high-resolution reanalysis data and measurements from over 1,400 ground monitoring stations, TopoFlow achieves a PM2.5 prediction RMSE of 9.71 μg/m³, outperforming operational forecasting systems by 71–80% and surpassing state-of-the-art AI models by 13%. The model maintains high accuracy across a 12–96 hour forecast window for four major pollutants, with errors substantially below China’s 24-hour air quality standard limits.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Neural Architectures and Foundation Models
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

We propose TopoFlow (Topography-aware pollutant Flow learning), a physics-guided neural network for efficient, high-resolution air quality prediction. To explicitly embed physical processes into the learning framework, we identify two critical factors governing pollutant dynamics: topography and wind direction. Complex terrain can channel, block, and trap pollutants, while wind acts as a primary driver of their transport and dispersion. Building on these insights, TopoFlow leverages a vision transformer architecture with two novel mechanisms: topography-aware attention, which explicitly models terrain-induced flow patterns, and wind-guided patch reordering, which aligns spatial representations with prevailing wind directions. Trained on six years of high-resolution reanalysis data assimilating observations from over 1,400 surface monitoring stations across China, TopoFlow achieves a PM2.5 RMSE of 9.71 ug/m3, representing a 71-80% improvement over operational forecasting systems and a 13% improvement over state-of-the-art AI baselines. Forecast errors remain well below China's 24-hour air quality threshold of 75 ug/m3 (GB 3095-2012), enabling reliable discrimination between clean and polluted conditions. These performance gains are consistent across all four major pollutants and forecast lead times from 12 to 96 hours, demonstrating that principled integration of physical knowledge into neural networks can fundamentally advance air quality prediction.

Problem

• air quality prediction
• pollutant dynamics
• topography
• wind direction
• high-resolution modeling

Innovation

• physics-guided neural networks
• topography-aware attention
• wind-guided patch reordering
• high-resolution air quality prediction
• vision transformer