Organic Speciation Related to Source Receptor Modeling

Dr. Eric Fujita, Desert Research Institute
Dr. Tad Kleindienst, U.S. Environmental Protection Agency
Dr. Tim Larson, University of Washington

What organic compounds (or compound classes) are useful for source apportionment?

·        Source composition of combustion sources varies greatly with fuel type, emission controls, environmental conditions, operating mode, and the methods and procedures for sample collection and chemical analysis. These factors must be carefully considered before applying available profiles in receptor and source modeling.

·        Most gasoline vehicles are relatively clean, especially in hot-stabilized mode. Virtually all of the PM emissions from “normal emitters” result from cold starts and hard accelerations with relatively higher amounts of elemental carbon.

·        A relatively small fraction of high emitters are responsible for a disproportionate amount of the total PM emissions from gasoline vehicles. They have cumulative PM emissions that are more linear with time than normal emitters with higher relative amounts of OC.

·        Elemental carbon is dominant in diesel exhaust, but is lower in newer technology diesel engines and during lower engine loads. Because of the variability of OC/EC splits, gasoline and diesel vehicles cannot be apportioned by carbon analysis alone, and EC is not a unique tracer for diesel exhaust.

·        Gasoline vehicle, whether low or high emitter, emit greater relative amounts of high molecular-weight particulate PAHs (e.g., benzo(b+j+k)fluoranthene, benzo(ghi)perylene, ideno(1,2,3-cd)pyrene, and coronene) that diesel vehicles. These PAHs are found in used gasoline motor oil but not in fresh oil nor in diesel engine oil. Diesel emissions contained higher proportions of dimethylnaphthalenes, methyl- and dimethylphenanthrenes, and methylfluorenes.

·        Gasoline vehicles emit volatile PAH’s (e.g., naphthalene and methylnaphthalenes) in amounts per unit of fuel that equals or exceeds that of diesel vehicles even “normal’ emitters. These compounds require back-up traps to be quantitatively collected. The potential contributions of semi-volatile PAH and alkanes from motor vehicles to formation of SOA have not been determined.

·        Hopanes and steranes are present in lubricating oil with similar composition for both gasoline and diesel vehicles and are not present in gasoline or diesel fuels. The relative abundances of hopanes and steranes to emissions of elemental carbon differ substantially for the diesel and gasoline vehicles. This difference in ratios of hopanes plus steranes to elemental carbon could be used to quantify the contribution of gasoline-powered and diesel-powered vehicles.

·        A wide range of volatile, semi-volatile and particulate organic compounds is emitted from wood combustion from the release of resinous compounds (e.g., retene and 1,7-dimethylphenanthrene), and decomposition of cellulose (levoglucosan), hemicelluloses and lignin (e.g., guaiacols, syringols and their derivatives). Reactivity and phase distribution of methoxyphenols must be considered in receptor modeling applications. Levoglucosan is chemically stable but is emitted by different vegetation in widely varying proportion to total PM. 

·        Certain fatty acids (e.g. palmitic acid, stearic acid and oleic acid) as well as cholesterol have been used as possible marker for meat smoke. Long-chain γ-lactones are formed by lactonization of β-hydroxy fatty acids normally found in triacylglycerols. They also result from the oxidation of alkenals and oleic acid. These compounds are emitted in small amounts relative to PM_2.5, but may be useful molecular markers for meat cooking.

·        The apparent contributions of SOA to ambient PM_2.5 concentrations have been associated in receptor modeling to the unidentified organic component. Both primary and secondary components of organic aerosols are complex and may partition between the gas and particulate phases with relative amounts that vary by region and season. Since there are no direct measurements for these two categories, their relative fractions are uncertain.

·        Aromatic and biogenic hydrocarbons are known to form SOA in laboratory smog chamber systems.  Such data can be useful for prospective modeling of SOA in ambient fine particulate matter using emission inventories of biogenic and aromatic hydrocarbons as inputs to an air quality model. 

·        The origin of SOA in ambient samples can be determined by examining the organic fraction of ambient PM_2.5 for marker compounds specific to various organic precursors or classes of precursors. Classes of compounds detected in ambient air include     dicarboxylic acids, triols and tetrols, (di) hydroxy acids, methoxy dicarboxylic acids, ketodicarboxylic acids, acyl dicarboxylic acids, tricarboxylic acids, oxoacids, carbonyl compounds, hydroxy carbonyl compounds, polyketones, and hydroxy polyketones.

·        Gas-phase PAH reactions produce nitro-PAHs and nitro-PAH lactones which may then condense onto particles.

·        The specific nitro-isomers formed can be used as “markers” of OH radical or of NO₃ radical chemistry in an airmass. Particle-associated markers of OH radical chemistry include 2-NF, 2-NP, 2-NBz and  high 2-NF/2-NP ratios are markers of NO₃ radical chemistry.

·        The ambient concentrations of phthalic acid correlate with 3-nitrobiphenyl, a marker of OH radical reaction. Phthalic acid and methyphthalic acids may be ultimate particle-associated products of naphthalene and methylnaphthalene reactions.    

·        The complexity of SOA formation is such that it is unlikely the use of tracer compounds will produce the level of results obtained for primary PM_2.5 emissions.

·        Multivariate receptor model algorithms (PMF, Unmix, COPREM, ME2) do not need source composition as inputs and simultaneously identifies “factors” and their contributions to a receptor sample. The ME2 algorithm can also include additional prior constraints (e.g. meteorology, particle size). Association between factors and source profiles can be ambiguous.

·        More work is needed to merge these newer receptor models with explicit descriptions of meteorological transport and atmospheric chemistry and methods are needed to better describe the uncertainties in model predictions.