Abstract
This study addresses the causal identification of air pollutants from surrounding cities affecting Beijing's air quality. A novel compressive sensing causality analysis (CS-Causality) method, which combines Granger causality analysis (GCA) and maximum correntropy criterion (MCC), is presented for efficient identification of the air pollutant causality between Beijing and surrounding cities. Firstly, taking the spatiotemporal correlation into consideration, the original data is mapped into low-dimensional space. Valid information is then obtained based on compressive sensing (CS), which can greatly reduce the dimensions of the data, thus decreasing the amount of data analysis required. Secondly, to analyze the causal relations, GCA, represented by the prediction from one time series to another, is extended to rule out “Non-Granger” causes of air pollutants in Beijing originating from its surrounding cities. Thirdly, the greatest impact on Beijing's air quality is confirmed based on MCC. Finally, the accuracy of these results is verified using the transfer entropy.
Introduction
In the era of "Urban Big Data," environmental monitoring has reached unprecedented levels of granularity. However, the sheer volume of data generated by thousands of sensors across multiple cities presents a significant challenge known as the "curse of dimensionality." For policymakers in Beijing, simply knowing that local air quality correlates with that of neighboring cities is insufficient. Effective intervention requires identifying causality: which specific cities are the sources of pollutants, the duration of the diffusion lag, and the magnitude of their impact. This paper introduces the CS-Causality framework to address these challenges by efficiently processing massive, noisy datasets to pinpoint causal environmental links.
The CS-Causality Framework
The proposed methodology consists of a three-stage pipeline designed to balance computational efficiency with analytical rigor:
- Compressive Sensing (CS) for Data Reduction: High-dimensional spatiotemporal data is mapped into a low-dimensional space. By leveraging the sparsity of the signal, CS removes redundant and noisy data points while preserving the essential information needed for reconstruction. This step significantly reduces the computational overhead for subsequent analysis.
- Extended Granger Causality Analysis (GCA): GCA is employed to determine if the historical pollutant levels of a surrounding city can improve the prediction of Beijing’s future air quality. This stage filters out "Non-Granger" causes, establishing a directed relationship between source and target cities.
- Maximum Correntropy Criterion (MCC): To quantify the impact, the researchers utilize MCC, a robust similarity measure. Unlike standard correlation metrics, MCC is highly effective at handling non-Gaussian noise, allowing for the identification of the "predominant" causal city for specific pollutants.
Experimental Results and Insights
The study analyzed daily averaged data for five major pollutants (PM2.5, PM10, SO2, NO2, and CO) from Beijing and five surrounding cities (Zhangjiakou, Chengde, Tianjin, Baoding, and Tangshan) during November 2019.
Efficiency and Stability
The researchers identified a critical performance threshold at a 65% compression ratio. At this level, the system achieved a 90% processing efficiency while maintaining a low error ratio of approximately 5%. Mathematical analysis further confirmed the mean-square stability of the weight-update algorithm used in the framework.
Causal Mapping and Diffusion Lag
The analysis revealed that air pollutants typically require 2 to 3 days to diffuse from surrounding regions into Beijing. More importantly, the causal sources were found to be pollutant-specific rather than uniform across all air quality metrics.
As illustrated in the causal maps above, Tangshan was identified as the primary contributor to Beijing’s PM2.5 and SO2 levels. Conversely, Baoding emerged as the leading driver for PM10 and NO2 concentrations. These findings were validated using Transfer Entropy (TE), showing high consistency between the CS-Causality model and non-parametric information transfer measures.
Conclusion
The CS-Causality framework provides a robust and efficient means of identifying the sources of urban air pollution. By integrating compressive sensing with advanced causality metrics, the model overcomes the limitations of high-dimensional data processing. Future work will focus on extending this framework to account for nonlinear atmospheric dynamics and integrating meteorological variables such as wind speed and humidity to further refine predictive accuracy.
Summary updated