DSGNN: A Dual-View Supergrid-Aware Graph Neural Network for Regional Air Quality Estimation

TL;DR

DSGNN addresses regional air quality estimation in regions without monitoring stations by introducing a dual-view, supergrid-aware graph neural network. It constructs AOD- and meteorology-informed dual-view supergrids, learns implicit correlations between supergrids, and fuses information from both views through a specialized message passing framework, enabling modeling of both proximal and distant spatial dependencies. Empirical results on two real-world datasets show state-of-the-art MAE gains, driven by the combination of dynamic/static supergrids, low-rank assignment learning, and cross-view fusion. The approach provides a scalable, data-driven means to estimate air quality with enhanced spatial coverage and interpretability, with potential extensions to forecasting and multi-level spatial modeling.

Abstract

Air quality estimation can provide air quality for target regions without air quality stations, which is useful for the public. Existing air quality estimation methods divide the study area into disjointed grid regions, and apply 2D convolution to model the spatial dependencies of adjacent grid regions based on the first law of geography, failing to model the spatial dependencies of distant grid regions. To this end, we propose a Dual-view Supergrid-aware Graph Neural Network (DSGNN) for regional air quality estimation, which can model the spatial dependencies of distant grid regions from dual views (i.e., satellite-derived aerosol optical depth (AOD) and meteorology). Specifically, images are utilized to represent the regional data (i.e., AOD data and meteorology data). The dual-view supergrid learning module is introduced to generate supergrids in a parameterized way. Based on the dual-view supergrids, the dual-view implicit correlation encoding module is introduced to learn the correlations between pairwise supergrids. In addition, the dual-view message passing network is introduced to implement the information interaction on the supergrid graphs and images. Extensive experiments on two real-world datasets demonstrate that DSGNN achieves the state-of-the-art performances on the air quality estimation task, outperforming the best baseline by an average of 19.64% in MAE.

Figures (8)

• Figure 1: Example of grid regions. (Best viewed in color).
• Figure 2: Framework of DSGNN. TE denotes the temporal encoder. LRM denotes the low-rank-based mapper. SGC denotes supergrid graph convolution. S2G update denotes grid region representation updating based on supergrid representation. G update denotes grid region representation updating. $\ominus$ denotes concatenation. $\otimes$ denotes multiplying. (Best viewed in color).
• Figure 3: Architecture of the temporal encoder.
• Figure 4: Visualization of the study areas. The study areas are in the red boxes. (Best viewed in color).
• Figure 5: Impact of hyper-parameters.