Original scientific paper
SOURCE APPORTIONMENT OF PM2.5 IN URBAN AREAS USING MULTIPLE LINEAR REGRESSION AS AN INVERSE MODELLING TECHNIQUE
Bruce Denby
; Norwegian institute for air research (NILU), Kjeller, Norway
Abstract
In many countries emissions of particulate matter from urban sources, such as traffic and domestic wood burning, can
lead to high episodic concentrations. Though it is important for air quality management and exposure studies to understand the
individual source contributions to these concentrations, the complexity of the urban environment does not always allow a clear
separation of sources when using conventional monitoring techniques that measure particulate mass only. Chemical analysis of the
particulates, combined with receptor modelling, is one method for determining source contributions but these do not provide direct
information on emissions. Inverse modelling methods, that make use of both dispersion models and measurements, can in principle
be used to determine emissions strengths and distributions. However, the urban environment is generally so complex and the
number of observations so limited that most inverse modelling methods cannot be effectively applied. In this paper a straight
forward inverse modelling method, using multiple linear regression, is described and applied. The method determines the optimal fit
of the calculated source contributions using dispersion modelling, providing scaling factors for the individual source contributions.
The method is applied to the urban area of Oslo for PM2.5 in the winter of 2004 and the results of the inverse modelling are
compared to independent receptor modelling. The method shows that the modelled source contribution from suspended road dust is
underestimated by a factor of 7 – 10. For domestic wood burning the method shows an overestimate of the modelled source
contribution by a factor of 2 - 3. These results are confirmed using independent analysis by receptor modelling. The methodology
cannot distinguish directly between model or emission error and so further assessment of the model itself, and its uncertainty, is
required before concrete statements concerning emission strengths can be made.
Keywords
Dispersion modelling; receptor modelling; multiple linear regression; particulate matter; inverse modelling; urban air quality; emissions; source apportionment
Hrčak ID:
64080
URI
Publication date:
12.12.2008.
Visits: 1.335 *