【摘要】 基于石家庄市2014年—2016年空气污染物观测数据,建立时间序列滑动窗,对窗体内主要空气污染物浓度的波动计算皮尔逊相关系数,根据其相关性进行粗粒化处理并得到相关关系模态,再以模态为节点,以窗体滑动次序连边,构建石家庄市空气污染物波动相关性时变网络,通过研究网络拓扑性质可得到以下结论:石家庄市PM2.5、CO、NO2、SO2的浓度波动互为正相关出现频次较高;空气污染物浓度波动相关性转化往往以全局相关为媒介,个别指标浓度与其他指标浓度的关联性是从不相关过渡到正相关;主要空气污染物浓度的当天波动相关性倾向于维持前一天波动相关性状态;空气污染物浓度波动相关性的转化会随季节变化而不同。更多还原

【Abstract】 Based on the observation data of air pollutants in Shijiazhuang city from 2014 to 2016, the time series sliding window is established first, then the Pearson correlation coefficient is calculated for the fluctuation of the main air pollutant concentration in the window, and the coarse graining treatment is carried out according to the correlation. Then the modality is used as a node, and the sliding order of the form is connected, to construct a timevarying network of air pollutant fluctuation correlation in Shijiazhuang city. By studying the network topology properties, the following conclusions can be drawn: The concentration fluctuations of PM2.5, CO, NO2 and SO2 in Shijiazhuang are positive relevant, and its frequency of occurrence is higher. For a long time, the correlation conversion of the main air pollutant concentration fluctuations is often based on the global correlation. In a short time, the correlation between the concentration of individual indicators and the concentration of other indicators is from irrelevant to positive correlation. The day-to-day fluctuation correlation of the main air pollutant concentration tends to maintain the fluctuation correlation state of the previous day. The conversion of the correlation of fluctuations in air pollutant concentration will vary with the seasons.更多还原