The increase in atmospheric pollution dominated by particles with an aerodynamic diameter smaller than 2.5 μm (PM_2.5) has become one of the most serious environmental hazards worldwide. The geographically weighted regression (GWR) model is a vital method to estimate the spatial distribution of the ground-level PM_2.5 concentration. Wind information reflects the directional dependence of the spatial distribution, which can be abstracted as a combination of spatial and directional non-stationarity components. In this paper, a GWR model considering directional non-stationarity (GDWR) is proposed. To assess the efficacy of our method, monthly PM_2.5 concentration estimation was carried out as a case study from March 2015 to February 2016 in the Yangtze River Delta region. The results indicate that the GDWR model attained the best fitting effect (0.79) and the smallest error fluctuation, the ordinary least squares (OLS) (0.589) fitting effect was the worst, and the GWR (0.72) and directionally weighted regression (DWR) (0.74) fitting effects were moderate. A non-stationarity hypothesis test was performed to confirm directional non-stationarity. The distribution of the PM_2.5 concentration in the Yangtze River Delta is also discussed here.

Prediction of traffic violations plays a key role in transportation safety. Combining with deep learning to predict traffic violations has become a new development trend. However, existing methods are based on regular spatial grids which leads to a fuzzy spatial expression and ignores the strong correlation between traffic violations and road network. A spatial topological graph can express the spatiotemporal correlation more accurately and then improve the accuracy of traffic violation prediction. Therefore, we propose a GATR (graph attention network based on road network) model to predict the spatiotemporal distribution of traffic violations, which adopts a graph attention network model combined with historical traffic violation features, external environmental features, and urban functional features. Experiments show that the GATR model can express the spatiotemporal distribution pattern of traffic violations more clearly and has higher prediction accuracy (RMSE = 1.7078) than Conv-LSTM (RMSE = 1.9180). The verification of the GATR model based on GNN Explainer shows the subgraph of the road network and the influence degree of features, which proves GATR is reasonable. GATR can provide an important reference for prevention and control of traffic violations and improve traffic safety.