PCDCNet: A Surrogate Model for Air Quality Forecasting with Physical-Chemical Dynamics and Constraints

2025-05-26

AI Summary

Air quality forecasting (AQF) faces challenges including complex physical mechanisms, high computational cost of numerical models, and poor generalizability of deep learning approaches. To address these, we propose a physics-constrained graph neural network (GNN) surrogate model that—uniquely—integrates differentiable atmospheric chemical kinetics equations into a GNN (for spatial transport modeling) and a gated recurrent unit (for temporal accumulation). The framework further incorporates multi-scale feature enhancement and a physics-consistency loss. By explicitly embedding emission sources, meteorological drivers, and nonlinear chemical transformations, the model achieves high-accuracy, station-level 72-hour forecasts of PM₂.₅ and O₃ across multiple cities: MAE improves by 18.7% and 15.2%, respectively. Inference is over 2,000× faster than CMAQ, while maintaining interpretability, cross-city generalizability, and low computational overhead. The model has been deployed in an operational public-service real-time forecasting platform.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Planning, Routing, and Scheduling: Planning with Language Models
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Air quality forecasting (AQF) is critical for public health and environmental management, yet remains challenging due to the complex interplay of emissions, meteorology, and chemical transformations. Traditional numerical models, such as CMAQ and WRF-Chem, provide physically grounded simulations but are computationally expensive and rely on uncertain emission inventories. Deep learning models, while computationally efficient, often struggle with generalization due to their lack of physical constraints. To bridge this gap, we propose PCDCNet, a surrogate model that integrates numerical modeling principles with deep learning. PCDCNet explicitly incorporates emissions, meteorological influences, and domain-informed constraints to model pollutant formation, transport, and dissipation. By combining graph-based spatial transport modeling, recurrent structures for temporal accumulation, and representation enhancement for local interactions, PCDCNet achieves state-of-the-art (SOTA) performance in 72-hour station-level PM2.5 and O3 forecasting while significantly reducing computational costs. Furthermore, our model is deployed in an online platform, providing free, real-time air quality forecasts, demonstrating its scalability and societal impact. By aligning deep learning with physical consistency, PCDCNet offers a practical and interpretable solution for AQF, enabling informed decision-making for both personal and regulatory applications.

Problem

• Bridging gap between numerical models and deep learning for AQF
• Reducing computational costs while maintaining physical-chemical accuracy
• Improving generalization in air quality forecasting with physical constraints

Innovation

• Integrates numerical modeling with deep learning
• Uses graph-based spatial transport modeling
• Deploys online for real-time air quality forecasts