Mixed multiway analysis of airborne particle composition data

Airborne particle composition data were obtained from week-long samples col lected at the northernmost manned site in the world, Alert, Northwest Terri tories, Canada, during the period from 1980 to 1991. It was found that the measured weekly average concentrations display strong persistent seasonal v ariations. Initially the measured concentrations of 24 constituents were ar ranged into both two-way and three-way data arrays, and bilinear and trilin ear models were used to fit the data using positive matrix factorization (P MF). Five factors were found to explain the data quite well for both two-wa y and three-way modeling, and each factor represented a likely particle sou rce. In the two-way modeling the yearly cyclical seasonal variations were n ot directly retrieved, since the whole 11 years of data were regarded as a single mode in the fitting. In the three-way analysis, fixed seasonality wa s imposed by assuming the week-to-week patterns of the source contributions to recur from year to year. The factors represent winter Arctic haze, phot ochemical sulfate after polar sunrise, biogenic sulfur, soil and sea salt. The resulting fit for some elements became worse because the year-to-year v ariation is not identical for these sources. These results suggested that a mixed two-way and three-way model might be the best representation of the data. The methodology to calculate such a mixed model has just been develop ed, namely the multilinear engine (ME). In this study the ME has been used to estimate a mixed two-way/three-way model for the Alert aerosol data. Fiv e two-way and two three-way factors have been found to provide the best fit and interpretation of the data. Each factor represents a probable source w ith a distinctive compositional profile and seasonal variations. The five t wo-way factors are (i) winter Arctic haze dominated by SO42 together with m etallic species and peaking from December to March, (ii) soil represented b y Si, Al and Ca, (iii) sea salt, (iv) sulfate with high acidity peaking in late March/April, and (v) iodine representing most of the observed I with t wo maxima, one around September/October and the other around March/April. T he two three-way factors are (i) bromine characterized by a maximum in the spring around March/April, and (ii) biogenic sulfur which includes sulfate and methane sulfonate (MSA) with maxima in May and August. The results obta ined are consistent with those obtained in the previous study and agree wit h the current understanding of the Arctic aerosol. In both analyses the yea r-to-year strength of the biogenic factor appears to correlate strongly wit h the average temperature in the northern hemisphere. This result suggests that as the temperature rises, there is increased biogenic production of th e reduced sulfur precursor compounds that are oxidized in the atmosphere to sulfate and MSA and could be evidence of a negative feedback mechanism in the global climate system that had been previously postulated. Copyright (C ) 1999 John Wiley & Sons, Ltd.