Double-Diffusion: Diffusion Conditioned Diffusion Probabilistic Model For Air Quality Prediction

Hanlin Dong, Arian Prabowo, Hao Xue, Flora D. Salim

Abstract

Air quality prediction is a challenging forecasting task due to its spatio-temporal complexity and inherent dynamics. While current models utilize Graph Neural Networks or physics principles, balancing deterministic certainty with stochastic uncertainty remains difficult. We propose Double-Diffusion, a novel diffusion probabilistic model that harnesses known physics to guide air quality forecasting. By integrating a physics-based conditional generative approach with a sampling strategy adapted from image restoration and a new denoiser architecture, Double-Diffusion achieves state-of-the-art results. It reduces inference time by 30% to 50% while improving the Continuous Ranked Probabilistic Score (CRPS) by 3-12% across real-world datasets.

Introduction

Predicting urban air quality, such as PM2.5 levels, is notoriously difficult due to two primary factors: spatio-temporal complexity (pollutant movement across urban grids) and inherent uncertainty (unpredictable shifts in traffic or weather).

Existing solutions typically fall into two categories. Deterministic models, like Physics-Informed Neural Networks (PINNs), capture patterns well but struggle with uncertainty quantification. Conversely, probabilistic models like Denoising Diffusion Probabilistic Models (DDPMs) excel at uncertainty but are computationally expensive and often ignore established physical laws. Double-Diffusion bridges this gap by using physics as a "shortcut" to guide the generative process, making it both more accurate and significantly faster.

Methodology

The "Double-Diffusion" framework refers to the synergy between physical diffusion (pollutant spreading) and probabilistic diffusion (the generative model).

1. Physics-Based Preliminary Forecast

Instead of initializing from pure noise, the model employs a Pollutant Transport Diffusion ODE based on Fick’s Law. This parameter-free step uses historical sensor data and the city's spatial graph to generate a rough physical estimate of pollutant spread.

2. Residual Learning (Resfusion)

Borrowing from image restoration, the model treats the physical forecast as a "degraded" signal. During training, it learns to predict "residual noise"—the specific difference between the physical ODE's output and the ground truth. This allows the model to focus specifically on real-world deviations that simple physics cannot account for.

The Double-Diffusion framework integrates a physics-based ODE solver with a residual-based diffusion process to refine air quality predictions.

3. The Zigzag Denoiser

To process these complex sequences, the authors introduced the Zigzag Denoiser. This architecture features:

  • Temporal Scaling: It upscales temporal resolution to capture fine-grained features before downscaling, unlike standard U-Nets.
  • gMLP Layers: By utilizing Spatial Gating Units (gMLPs) instead of heavy Transformers, the model captures long-range dependencies with significantly lower memory and compute overhead.
The Zigzag Denoiser architecture featuring temporal scaling, gMLP layers, and Graph Convolutional Networks (GCN).

Evaluation and Results

The model was tested on two major datasets: Beijing (KDD Cup 2018) and Guangzhou (UrbanAir), predicting the next 12 hours of PM2.5 levels based on a 12-hour history.

  • Efficiency: Double-Diffusion is 37% faster than DiffSTG and 53% faster than CSDI. By starting closer to the ground truth via physics guidance, it can skip redundant early diffusion steps.
  • Accuracy: The model ranked first in RMSE and CRPS across both datasets. The 3-12% improvement in CRPS indicates that its probability distributions are more reliable for real-world decision-making.
  • Robustness: Case studies showed the model successfully captured sudden pollution spikes that pure ODE-based models missed, demonstrating the power of its hybrid approach.

Conclusion

Double-Diffusion demonstrates that physics should not just be a constraint but can serve as a powerful condition for generative models. While currently limited by a simplified diffusion-only ODE and a focus on PM2.5 readings, future iterations could incorporate advection (wind) and auxiliary data like traffic volume to further refine urban air quality forecasting.

Summary updated

Abstract

Air quality prediction is a challenging forecasting task due to its spatio-temporal complexity and the inherent dynamics as well as uncertainty. Most of the current models handle these two challenges by applying Graph Neural Networks or known physics principles, and quantifying stochasticity through probabilistic networks like Diffusion models. Nevertheless, finding the right balancing point between the certainties and uncertainties remains an open question. Therefore, we propose Double-Diffusion, a novel diffusion probabilistic model that harnesses the power of known physics to guide air quality forecasting with stochasticity. To the best of our knowledge, while precedents have been made of using conditional diffusion models to predict air pollution, this is the first attempt to use physics as a conditional generative approach for air quality prediction. Along with a sampling strategy adopted from image restoration and a new denoiser architecture, Double-Diffusion ranks first in most evaluation scenarios across two real-life datasets compared with other probabilistic models, it also cuts inference time by 50% to 30% while enjoying an increase between 3-12% in Continuous Ranked Probabilistic Score (CRPS).
BibTeX
@misc{dong2025double,
  author = {Hanlin Dong and Arian Prabowo and Hao Xue and Flora D. Salim},
  title = {{Double-Diffusion: Diffusion Conditioned Diffusion Probabilistic Model For Air Quality Prediction}},
  year = {2025},
  eprint = {2506.23053},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2506.23053},
}
Figure