Analytica Chimica Acta 677 (2010) 64–71 Contents lists available at ScienceDirect Analytica Chimica Acta journal homepage: www.elsevier.com/locate/aca Comparing roadsoils pollution patterns extracted by MOLMAP and classical three-way decomposition methods M.P. Gómez-Carracedo a , D. Ballabio b , J.M. Andrade a,∗ , J. Aires-de-Sousa c , V. Consonni b a Department of Analytical Chemistry, University of A Coru˜ na, Campus da Zapateira s/n, E-15071 A Coru˜ na, Spain b Milano Chemometrics and QSAR Research Group, Department of Environmental Sciences, University of Milano-Bicocca, P.za della Scienza, 1-20126 Milano, Italy c CQFB and REQUIMTE, Department of Chemistry, Faculdade de Ciências e Tecnologia, Universidade Nova de Lisboa, 2829-516 Caparica, Portugal article info Article history: Received 27 April 2010 Received in revised form 20 July 2010 Accepted 28 July 2010 Available online 6 August 2010 Keywords: Self-organizing maps Parallel factor analysis Matrix augmentation principal components analysis Procrustes rotation Soils Heavy metals abstract A recent approach based on self-organizing maps (SOMs) to extract patterns from three-way data, named MOLMAP, was applied in a four-seasons study on soil pollution and its results compared with three different conventional approaches: Parallel factor analysis (PARAFAC), matrix augmented principal com- ponents analysis (MA-PCA) and Procrustes rotation. Each sampling season comprised 92 roadsoil samples and 12 analytical variables (Cd, Co, Cu, Cr, Fe, Mn, Ni, Pb, Zn, loss on ignition, pH and humidity). It was found that all techniques yielded highly similar results as the samples became organized in two major groups, each with a differentiated pollution pattern. This confirmed MOLMAP as a reliable option to handle environmental three-way datasets and to extract accurate pollution patterns. © 2010 Elsevier B.V. All rights reserved. 1. Introduction Pollution by heavy metals over large geographical areas and long periods of time may cause chronic damage to living organisms and must be monitored carefully to determine the extent of the environmental contamination [1]. Road traffic contributes signifi- cantly to air pollution in urban areas, generating particulate matter, aerosols and heavy metals around roads [2,3]. Deposition of those heavy metals on vegetation, cultivated crops and soil may influ- ence the elemental composition and physiology of leaves [1,3] and, ultimately, the health of animals and human beings. Therefore, soil pollution monitoring is an important issue in today’s environmen- tal protection and remediation. The analysis and monitoring of soils have been growing steadily as many European cities set environmental monitoring programs to assess the quality of their natural surroundings (including water, air and soil). New developments in analytical methodologies allowed for incredibly low limits of detection, large sensitivity and spe- ciation capabilities. Nevertheless, the analytical workhorse needs close linkage with chemometric tools to extract maximum relevant information from the bulk of data obtained after an environmental study. In particular, there is an ongoing interest in studying simul- ∗ Corresponding author. Fax: +34 981167065. E-mail address: andrade@udc.es (J.M. Andrade). taneously the sets of data generated in different sampling periods from the same study region. Such data can be, typically, arranged as a three-way data cube (samples × variables × sampling periods). Important efforts have been reported in the last years in the environmental sciences to study those data cubes (in general, N- way data arrangements). Just as a matter of example, Felipe-Sotelo et al. applied parallel factor analysis (PARAFAC) and matrix aug- mentation principal components analysis (MA-PCA) to assess the temporal evolution of a river water quality [4]. Pardo et al. [5] employed classical two-way principal components analysis (PCA) and three-way analyses (MA-PCA, 3-PCA, PARAFAC and Tucker3) to study metals in soils. Singh et al. [6] analyzed a multiway dataset of different particle sizes of river bed sediments to evaluate the associ- ations of heavy metals and their chemical fractions with the particle sizes and their spatial distribution employing Tucker3. Andrade et al. used three different approaches for three-way analyses (Pro- crustes rotation, PARAFAC and MA-PCA) to study soil pollution [7] and oil spillages [8]. Finally, Tauler et al. [9] compared four receptor modelling techniques to perform source apportionment of physic- ochemical variables in airborne samples. Among the clustering techniques self-organizing maps (SOMs), or Kohonen Neural Networks, have shown advantageous capabili- ties to extract sample patterns in a non-hierarchical, unsupervised way [10–12]. Recently, the combination of SOMs and objective vari- able selection by CART (classification and regression trees) has been proposed as an effective mean to accelerate the optimization of 0003-2670/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.aca.2010.07.044