Air quality expert group - Fine particulate matter (PM2.5) in ... - Defra
Air quality expert group - Fine particulate matter (PM2.5) in ... - Defra
Air quality expert group - Fine particulate matter (PM2.5) in ... - Defra
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
<strong>PM2.5</strong> emissions and receptor modell<strong>in</strong>g<br />
63. This list can be used to guide future areas of research so that modellers<br />
have access to <strong>in</strong>formation that goes beyond what is provided by traditional<br />
<strong>in</strong>ventories.<br />
4.5 Receptor modell<strong>in</strong>g to estimate the source<br />
apportionment of <strong>PM2.5</strong><br />
64. Receptor modell<strong>in</strong>g refers to the use of monitor<strong>in</strong>g data collected <strong>in</strong> the<br />
atmosphere (as opposed to the modell<strong>in</strong>g of stack emissions) to <strong>in</strong>fer the<br />
sources responsible for the measured concentrations of a pollutant. In many<br />
situations it can yield quantitative as well as qualitative estimates. Two<br />
generic methods are used most commonly for receptor modell<strong>in</strong>g of airborne<br />
concentrations. Both require the collection of temporally-resolved, chemicallyspeciated<br />
data on the composition of airborne particles, often supplemented, <strong>in</strong><br />
the case of the multivariate statistical method, by meteorological and gas phase<br />
pollutant data. The two types of method are:<br />
(a) Chemical mass balance (CMB). This method requires a priori knowledge<br />
of the composition of all sources contribut<strong>in</strong>g to the airborne pollution,<br />
but not their emission rates. The measured air <strong>quality</strong> is assumed to<br />
be a l<strong>in</strong>ear sum of the contributions of the known sources, which are<br />
summed over each different sampl<strong>in</strong>g period to give the best match<br />
to the concentrations of the many chemical species measured <strong>in</strong> the<br />
atmosphere. In many studies, organic “molecular markers” which may<br />
be only m<strong>in</strong>or constituents of emissions are measured, as these help to<br />
discrim<strong>in</strong>ate between similar sources (e.g. petrol and diesel eng<strong>in</strong>es).<br />
This method has been applied to airborne particles sampled <strong>in</strong> the West<br />
Midlands (Y<strong>in</strong> et al., 2010).<br />
(b) Multivariate statistical methods. A suite of methods is based upon<br />
factor analysis, of which Positive Matrix Factorisation (PMF) has been<br />
developed specifically for the purpose of source apportionment of<br />
air <strong>quality</strong> data, and is the most commonly applied. Earlier studies<br />
used Pr<strong>in</strong>cipal Component Analysis, but PMF has the advantages of<br />
be<strong>in</strong>g constra<strong>in</strong>ed not to give negative solutions, and allow<strong>in</strong>g the<br />
weight<strong>in</strong>g of <strong>in</strong>put variables accord<strong>in</strong>g to analytical uncerta<strong>in</strong>ty. The<br />
method requires no a priori knowledge of source composition, but<br />
such data are valuable <strong>in</strong> discrim<strong>in</strong>at<strong>in</strong>g between similar sources. The<br />
method requires a substantial number of separate air samples (at least<br />
50) which are analysed for a wide range of chemical constituents.<br />
Constituents which come from the same source have the same temporal<br />
variation and if unique to that source are perfectly correlated. Typically,<br />
however, a given chemical constituent will have multiple sources and<br />
the programme is able to view correlations <strong>in</strong> a multidimensional<br />
space and can generate chemical profiles of “factors” with a unique<br />
temporal profile characteristic of a source. Past knowledge of source<br />
chemical profiles is used to assign factors to sources; typically up to<br />
ten different sources can be assigned factors. The method works best<br />
with a large dataset <strong>in</strong> which the number of samples far exceeds the<br />
number of analytical variables, and gives a clearer dist<strong>in</strong>ction of sources<br />
if sampl<strong>in</strong>g times are short, so that overlap of multiple po<strong>in</strong>t source<br />
contributions to a given sample is m<strong>in</strong>imised. Inclusion of meteorological<br />
101