| Popis: |
This paper demonstrates solutions to some of the problems normally encountered in the analysis of atmospheric chemical data sets. Multivariate data analysis techniques were applied to a unique chemical data set consisting of rainwater trace element concentrations determined via instrumental neutron activation analysis (INAA). Experimental uncertainties, principal component analysis (PCA) and principal component classification were used to categorize sources of variability in the data while partial-least-squares regression (PLS) was used to compare the rainwater data to crustal, sea salt, and smelter “signatures” of the same trace elements. These three source contributions to the rainwater trace element data were calculated and four other contributions from derived sources were interpreted based upon their chemical composition, geographical distribution, and temporal variation. |
