大气污染物监测数据不确定度评估方法体系建立及其对PMF源解析的影响分析

Establishment of an uncertainty assessment framework for atmospheric pollutant monitoring data and its impact on PMF source apportionment

目前,正定矩阵因子法(PMF)在大气污染物来源解析中得到了广泛应用,而监测数据不确定度评估是PMF源解析的重要内容之一.当前绝大部分研究采用的数据不确定度仅通过借鉴前人的方法简单计算而得,缺乏对计算方法的合理性和适用性进行评估.本研究通过3种常用的不确定度算法交互应用建立了一套有效评估不确定度的方法体系.通过对香港荃湾站点1998—2008年PM_(10)组分监测数据进行源解析并与平行采样数据源解析结果进行对比,对该方法体系的合理性进行了验证.结果发现,应用该方法体系可以将某些常规方法无法分离的因子进一步分解,得到的源解析结果的残差值甚至小于平行采样方法解析结果的残差值,得到的因子贡献率均处于常规方法得出的因子贡献率之间.这些均表明了该方法体系所得源解析结果的可靠性和全面性.因此,本研究建立的评估不确定度的方法体系具有较强的可行性,对确保源解析结果的准确性有重要意义.

Positive matrix factorization(PMF)has been applied extensively in source apportionment of atmospheric pollutants.Data uncertainty is one of the most important input information for PMF analysis.Currently,data uncertainty is calculated by some simple formulae referenced from previous studies,lacking evaluation of their reasonableness and applicability.In this study,we develop an uncertainty assessment framework(UAF)by utilizing three common uncertainty calculation algorithms.The effectiveness of this UAF is assessed by comparing with the source apportionment results on parallel measurement of PM10 compositions at Tsuen Wan station in Hong Kong during 1998-2008.It is found that applying UAF could further decompose some factors which couldn′t by traditional methods.The derived scaled residuals are even smaller than those from source apportionment on parallel measurement.Source contributions derived by UAF are well in between those from traditional methods.All above indicates the reliability and completeness of source apportionment by UAF.It is thus concluded that this UAF has great applicability and usefulness in ensuring the accuracy of source apportionment results.