Quantifying the scale-dependent relationships of PM and O on meteorological factors and their influencing factors in the Beijing-Tianjin-Hebei region and surrounding areas.

To investigate the variations of PM and O and their synergistic effects with influencing factors at different time scales, we employed Bayesian estimator of abrupt seasonal and trend change to analyze the nonlinear variation process of PM and O. Wavelet coherence and multiple wavelet coherence were utilized to quantify the coupling oscillation relationships of PM and O on single/multiple meteorological factors in the time-frequency domain. Furthermore, we combined this analysis with the partial wavelet coherence to quantitatively evaluate the influence of atmospheric teleconnection factors on the response relationships. The results obtained from this comprehensive analysis are as follows: (1) The seasonal component of PM exhibited a change point, which was most likely to occur in January 2017. The trend component showed a discontinuous decline and had a change point, which was most likely to appear in February 2017. The seasonal component of O did not exhibit a change point, while the trend component showed a discontinuous rise with two change points, which were most likely to occur in July 2018 and May 2017. (2) The phase and coherence relationships of PM and O on meteorological factors varied across different time scales. Stable phase relationships were observed on both small- and large-time scales, whereas no stable phase relationship was formed on medium scales. On all-time scales, sunshine duration was the best single variable for explaining PM variations and precipitation was the best single variable explaining O variations. When compared to single meteorological factors, the combination of multiple meteorological factors significantly improved the ability to explain variations in PM and O on small-time scales. (3) Atmospheric teleconnection factors were important driving factors affecting the response relationships of PM and O on meteorological factors and they had greater impact on the relationship at medium-time scales compared to small- and large-time scales.