非随机化人工增雨作业功效数值分析和效果评估

Numerical Analysis of Statistical Power in Precipitation Enhancement Experiment in Beijing and Estimation of Operational Cloud Seeding Effectiveness

用自然复随机化方法对北京市人工增雨作业非随机化试验进行功效数值分析。结果表明不同统计检验方案功效差别较大,序列试验功效最差,当作业样本数较多时对比试验功效较高,其次为区域历史回归试验和双比分析方案,当作业样本数较少时区域历史回归试验功效比其他两种方案更高。功效与增雨效果、历史样本数、作业样本数都有关系,当作业样本数或历史样本数增多时,功效都会增大,但是增大的程度会随着样本数目的增多而趋缓。分类统计不一定可以提高检验功效值,要采用分类统计方案首先就要保证分类后的作业样本数下的功效值大于合并后作业样本数下的功效值。在综合分析影响功效的各种因素基础上选择区域对比试验、双比分析和区域历史回归试验对北京市2002～2007年人工增雨作业进行效果总评价,结果均表明相对增雨效果在10%左右,采用复随机化方案进行显著度检验,结果表明3种方案下显著度水平均达0.05。

The naive re-randomization test is used for numerical analysis of the statistical power in a precipitation en- hancement experiment in Beijing. The results show that the statistical powers vary depending on the evaluation method; the sequence method provides the worst results. For a large number of samples, the comparison method has the highest statistical power, followed by the historical regression method and double ratio method. For a small number of samples, the historical regression method has a higher statistical power than the other two methods. The statistical power is strongly correlated with the precipitation enhancement effectiveness, historical samples, and operating samples. The sta- tistical power increases with increasing number of historical or operating samples, but the extent of the increase isreduced as the number of units increases. Classification will not always improve the detection efficiency; the statistical power of the post-classification sample size must be greater than that of the combined sample size if the classification method is used. Here, the comparison, double ratio, and historical regression methods are used to evaluate the precipita- tion enhancement effectiveness in Beijing from 2002 to 2007; the results of all three methods show that the relative effectiveness is about 10% at a significance of 0.05.