AAAR 32nd Annual Conference
September 30 - October 4, 2013
Oregon Convention Center
Portland, Oregon, USA
Abstract View
ME-2 Analysis of Long-term On-line Mass Spectrometric Data of Non-refractory Submicron Aerosol in the City of Zurich
FRANCESCO CANONACO, Jay Slowik, Urs Baltensperger, Andre Prévôt, Paul Scherrer Institute
Abstract Number: 315 Working Group: Source Apportionment
Abstract We present measurements of submicron aerosol composition from an aerosol chemical speciation monitor deployed in downtown Zurich. The ACSM yields quantitative mass spectra of the non-refractory aerosol component with an averaging period of 15 to 30 minutes. These are one of the first full-year measurements in Europe using this technique. The ACSM is capable of resolving inorganic species such as NH$_4, SO$_4, NO$_3, Cl, and the organics. The data can be further deconvolved in order to identify and apportion the sources, in particular for the complex organic fraction.
There are several possible techniques for the deconvolution of a data set. Positive matrix factorization (PMF) has been extensively exploited for this type of data. However, the multilinear engine 2 (ME2) offers advantages compared to PMF, due to the fact that the user has the possibility to introduce a priori information in form of e.g. known factor profiles into the model.
The aim of this work is to make use of the multiple approaches available in ME2 on the obtained data from downtown Zurich during the full year 2011. For this purpose we implemented a user-friendly interface in the software IGOR Pro.
We will discuss identification and source apportionment of ACSM organic data at an urban site over the full year of 2011 with the ME-2 algorithm. In particular, partially constrained source profiles lead to a successful separation of different sources including traffic wood burning, cooking and more which could not be apportioned by PMF. Especially interesting are the separation of several semi volatile secondary organic aerosol species that could be partially assigned to specific precursor groups. Different techniques, like the a-value or the pulling approach have been tested. Moreover, studying and comparing the outcomes of these just mentioned techniques provides useful information for estimating of the model uncertainty.