源解析中环境受体数据处理与扩展方法的建立

Method of the processing and expanding the environment receptor data in source apportionment

本文对包含污染源或气象条件等发生剧烈变化且数据量小的实测受体数据进行了正态扩展,得到扩展数据.同时,利用PMF和PCA模型验证了扩展方法得到的数据量是否能满足模型的要求.结果表明,扩展范围和扩展受体成分谱个数是两个影响扩展数据合理性的主要因素,最佳扩展条件为:扩展范围取标准差的0.5倍,扩展受体成分谱个数为6个.通过53 h算法标记出每个化学成分时间序列中能够代表污染源或气象条件等发生剧烈变化的值并给出对应的估计值.将与估计值的相对误差(RE)超过80%的被标记的值剔除,其余的替换成估计值,发现扩展后PCA解析结果与原始数据处理后数据解析结果基本一致,能够得到主要贡献的污染源及贡献率;若将被标记的值全部剔除后,则不适合做PCA解析;仅通过PMF验证,且扩展数据与原始数据的解析结果中污染源类的判断一致.将受体数据各化学成分的时间序列中代表污染源或气象条件等发生剧烈变化的值替换成53 h算法给出的估计值,对于受体数据量小且用无源成分谱的多元统计方法解析无法给出结果的情况下的源解析有更大价值.

The method of the normal distribution based data expansion was used to extend receptor data. The data for possessing the characteristic of tremendous change of source or meteorological condition were required to be expanded for applying to PMF and PCA models. The scope and the amount of expansion of the receptor data are the main influencing factors on the data expansion rationality,and the optimal scope was 0. 5 times of the standard deviation of each chemical composition and the amount of expansion was six respectively. The data for possessing the characteristic of tremendous change of source or meteorological condition were marked in time series of each chemical composition by 53 h algorithm. The estimation of marked data were also provided. If all marked data were removed,the remained data couldn＇ t meet the requirement of PCA model,but could be used for PMF model. The marked data with the relative error（ RE） values between original and estimated data being over 80% were removed. The remaining marked data were replaced by estimated data. Then,the expansion data was calculated by the K-means cluster and the normal expansion. The source apportionment results of data expansion consisted with that of original data basically by PCA model. This method had greater value in source apportionment for the less receptor data by using of the multivariate receptor models without source profiles.