DUGi

Source Apportionment of Atmospheric Trace Gases and Particulate Matter: Comparison of Log-ratio and Traditional Approaches

http://dugi.udg.edu/item/http:@@@@hdl.handle.net@@2072@@273641

In this paper we compare multivariate methods using both traditional approaches, whichignore issues of closure and provide relatively simple methods to deal with censored ormissing data, and log-ratio methods to determine the sources of trace constituents in theatmosphere. The data set examined was collected from April to July 2008 at a sampling sitenear Woods Hole, Massachusetts, along the northeastern United States Atlantic coastline.The data set consists of trace gas mixing ratios (O3, SO2, NOx, elemental mercury [Hgo], andreactive gaseous mercury [RGM]), and concentrations of trace elements in fine (<2.5 µm)particulate matter (Al, As, Ba, Ca, Cd, Ce, Co, Cs, Fe, Ga, Hg, K, La, Mg, Mn, Na, P, Pb,Rb, Sb, Sr, Th, Ti, V, Y, and Zn) with varying percentages of censored values for eachspecies. The data were separated into two subcompositions: s1, which is comprised by RGM andparticulate Hg (HgP), which are both highly censored; and s2 which includes all of the traceelements associated with particulate matter except Hg, and the trace gases O3, SO2, NOx, andHgo. Principal factor analysis (PFA) was successful in determining the primary sources forconstituents in s2 using both traditional and log-ratio approaches. Using the traditionalapproach, regression between factor scores and RGM and particulate Hg concentrationssuggested that none of the sources identified during PFA led to positive contributions ofeither reactive mercury compound. This finding is counter to most conventional thinking andis likely specious, resulting from removal of censored data (up to >80% of the entire dataset)during the analysis. Log-ratio approaches to find relationships between constituents comprising s2 with RGMand HgP (i.e., s1) focused on log-ratio correlation and regression analyses of alr-transformeddata, using Al as the divisor. Regression models accounted for large fractions of the variancein concentrations of the two reactive mercury species and generally agreed withconceptualizations about the formation and behavior of these species. An analysis ofindependence between the subcompositions demonstrated that the behavior of the twoconstituents comprising s1 (i.e., RGM and HgP) is dependent on changes in s2. Our findingssuggest that although problems related to closure are largely unknown or ignored in theatmospheric sciences, much insight can be gleaned from the application of log-ratio methodsto atmospheric chemistry data

Universitat de Girona. Departament d’Informàtica i Matemàtica Aplicada