Extreme haze was often observed at many locations in Beijing-Tianjin-Hebei region within several hours when they occurred, which is referred to as extreme co-movements and extreme dependence in statistics. This article applies tail quotient correlation coefficient to explore the temporal and spatial extreme dependence patterns of haze in this region. Hourly PM2.5 station-level data during 2014-2018 are used, and the results show that the tail quotient correlation coefficient between stations increases with month. Specifically, the simultaneous extreme dependence was strong in the fourth season, while the haze was severe. In the first season, while the haze was also severe, the extreme hazes only show strong co-movements with a time difference. These observations lead to the study of two special scenarios, that is, the concurrence/extreme dependence of the worst extreme haze and its lag effects. City clusters suffering simultaneous extreme haze or with certain time difference as well as the most frequently co-movement cities are identified. The extreme co-movements of these cities and the reasons for their occurrences have strong implications for improving the PM2.5 joint prevention and control in the Beijing-Tianjin-Hebei region. The importance of lag effects is also reflected in the precedence order of the extreme haze's appearance. It is especially useful when setting the mechanism of the early warning system which can be triggered by the first appearance of extreme haze. The precedence orders also avail in investigating the transmission path of the haze, based on which more precise meteorological models can be made to benefit the haze forecasting of the region.

中文翻译：

---

京津冀地区雾霾极端联动及其极端依赖模式识别.

京津冀地区多地往往在发生后数小时内就出现极端雾霾，统计上称之为极端联动、极端依赖。本文应用尾商相关系数探讨该地区雾霾的时空极端依赖性模式。采用2014-2018年逐小时PM2.5站级数据，结果表明，站间尾商相关系数随月份增大。具体来看，第四季同时极端依赖较强，雾霾严重。第一季，虽然雾霾也很严重，但极端雾霾只表现出强烈的时差联动。这些观察结果导致了对两种特殊情景的研究，即最严重的极端雾霾的并发/极端依赖性及其滞后效应。识别同时遭受极端雾霾或存在一定时差的城市群以及联动最频繁的城市。这些城市的极端联动及其发生的原因对于完善京津冀地区PM2.5联防联控具有重要意义。滞后效应的重要性还体现在极端雾霾出现的先后顺序上。在设置可在极端雾霾首次出现时触发的预警系统机制时尤其有用。优先顺序还有助于调查雾霾的传播路径，在此基础上可以建立更精确的气象模型，有利于该地区的雾霾预报。