کد مقاله | کد نشریه | سال انتشار | مقاله انگلیسی | نسخه تمام متن |
---|---|---|---|---|
4434953 | 1310536 | 2011 | 7 صفحه PDF | دانلود رایگان |

Uncertainty–weighted partial least squares discriminant analysis was used to identify key species that were subsequently included in the EPA CMB8.2 chemical mass balance model to assess PM2.5 source contributions from a previously published data set on school bus self–pollution. Estimates from this two– step modeling approach, herein referred to as effective variance discriminant analysis chemical mass balance (EVDA–CMB) were compared for eight separate runs with independent estimates from a synthetic tracer method. EVDA–CMB model predictions agreed favorably with those from the tracer method (R2 = 0.83, 0.96 and 0.48, for contributions from the bus tailpipe, the engine crankcase and from other sources, respectively). Predictions from the traditional CMB model (without prior species selection), did not agree as well with the tracer method estimates of the bus tailpipe and engine crankcase contributions (R2 = 0.18, 0.69, respectively), but did agree as well with the contributions from other sources (R2 = 0.60). Although this study required discrimination of only a few sources, the same approach could be applied to the more general receptor modeling problem as an initial screening procedure, including approaches that optimize the choice of variables based on ambient data. This is important given that the number of species available for use in receptor modeling is rapidly expanding.
Journal: Atmospheric Pollution Research - Volume 2, Issue 2, April 2011, Pages 144–150