基于人工影响天气技术的石河子冬季城市空气质量改善试验效果统计分析

采用2019年12月天山北部石河子（目标区）和昌吉（对比区）地区的空气质量指数（AQI）资料，对石河子人工改善空气质量的效果进行了分析，提供了基于人工影响天气技术消减AQI的证据.试验为期31 d，其中在13个试验日中进行了飞机作业.采用区域回归方法，建立两地AQI回归方程，计算目标区在未进行人工作业时的AQI期望值，将其与实测值对比，统计分析作业效果.试验期间，目标区与对比区AQI日变化呈高度的相关性，数据样本相关性经t检验满足要求，经F检验可信程度达到99%，显著性非常高.此外，基于两个随机试验验证统计结果的可靠性.结果表明，在石河子开展人工消雾作业期，AQI指数相较于不进行作业时平均减小52，相对减小28.1%.进一步进行90%置信区间估算结果表明，开展作业后目标区AQI平均值减小了23.71，相对减小15.1%.通过两个蒙特卡洛随机试验表明，人工作业对AQI产生的作用超过了试验期间大部分情况下两地AQI自然变化的区别，因此，统计结果有较高的可信度，在样本数足够时，人工作业的作用可以从复杂的自然变化中体现出来.但是，对于单日内的污染过程，其自然变化十分复杂，人工作业的作用难以进行定量化计算.

Statistical analysis of the effect of wintertime air quality improvement using weather modification technology in Shihezi

In this study, a statistical analysis is made to investigate the impact of artificial fog dispersal in Shihezi, north of Tianshan Mountain, using the data of air quality index (AQI) observed in Shihezi (target area) and Changji (control area) in Dec. 2019. The results provide statistical evidence that the AQI can be reduced by artificial fog dispersal. Data collected in 31 days are used, and in 13 of them airborne-based fog dispersal operation was conducted. A regression analysis is used to calculate the expected AQI without fog dispersal, and the expected AQI is compared to the observed AQI to estimate the impact of fog dispersal. During the measurement period, the AQI in the target and control areas are highly correlated. The t-test of the samples satisfies the requirement to establish a regression function, and the F-test indicates the confidence level is as high as 99%. In addition, two Monte Carlo random experiments are used to examine the robustness of the results. The analysis shows the average AQI in Shihezi decreased by 52 (28.1%) during the days with artificial fog dispersal. A further test based on 90% confidence level indicates the average AQI in Shihezi decreased by 23.71 (15.1%) during the days with artificial fog dispersal. Two Monte Carlo random experiments show the signal of artificial fog dispersal is higher than the natural variability in most situations, providing the confidence that the impact of fog dispersal can be implied if the sample size is large enough. However, for the fog process in a single day, the natural variability is often too large to tease out the impact of fog dispersal.