Statistical analysis of the concentrations of twelve metals in the Ligurian atmospheric aerosol

A correlation analysis based on the concept of singular-value decomposition of a matrix is applied here to an atmospheric aerosol data set, for twelve metals (Al, Ca, Cd, Co, Cu, Fe, Mg, Mn, Na, Ni, Pb and Zn),in the Ligurian Sea coastal environment. An iterative procedure allows atypical samples to be discarded, i.e. samples with characteristics significantly different to those of the whole data set. After five atypical samples have been objecti vely discarded, the general characteristics of the data set are better desc ribed and some features are enhanced according to a decreasing order of imp ortance, For example, the mineral dust character of almost every metal is s hown, which points out that the marine recycling component is negligible. M arine (Mg, Na) and anthropogenic (Cd, Cu, Pb, Zn) contributions are assesse d as well. There is also evidence of some peculiar behaviours: among others , Fe clearly exhibits an anthropogenic component; the remainder of automoti ve Pb is still apparent; Ni exhibits a relatively high crustal component; Z n, despite its recognized anthropogenic character, is not clearly correlate d with other anthropogenic metals. (C) 2000 Ifremer/CNRS/IRD/Editions scien tifiques et medicales Elsevier SAS.