Analytica Chimica Acta 688 (2011) 104–121

Contents lists available at ScienceDirect

Analytica Chimica Acta
journal homepage: www.elsevier.com/locate/aca

Review

The role of chemometrics in single and sequential extraction assays: A review Part I. Extraction procedures, uni- and bivariate techniques and multivariate variable reduction techniques for pattern recognition
Ornella Abollino ∗ , Mery Malandrino, Agnese Giacomino, Edoardo Mentasti
Department of Analytical Chemistry, University of Torino, Via Giuria 5, 10125 Torino, Italy

a r t i c l e

i n f o

a b s t r a c t
Element mobility and availability in natural solid matrices can be studied with single and sequential extraction procedures; such procedures provide reliable and useful information only if the experiments are correctly planned and executed and the results are properly interpreted. Chemometrics can be a valuable tool for these aims, especially taking into account the large amounts of data generated with extraction essays and the complexity of the processes under investigation. This review deals with the application of chemometrics in research studies involving single and sequential extractions on soils or sediments, for several purposes: the development and optimization of the extraction conditions, the calculation of element fractionation, the visual illustration of the experimental results, the acquisition of different areas of information, including relationships among variables, similarities and differences among samples, causes of the observed behaviour (e.g. source identification), risk assessment, models and predictions of future events. In Part I of the review, following an overview on extraction procedures, the applications of univariate and bivariate chemometric methods are reported; then the principles of multivariate techniques for pattern recognition based on variable reduction, their applications and the main findings obtained are addressed. © 2011 Elsevier B.V. All rights reserved.

Article history: Received 16 June 2010 Received in revised form 23 November 2010 Accepted 13 December 2010 Available online 6 January 2011 Keywords: Single extraction Sequential extraction Chemometrics Multivariate statistics Soil Sediment

Ornella Abollino received her PhD in Chemical Sciences in 1991 from the University of Torino. She is presently Associate Professor of Analytical Chemistry at the Faculty of Pharmacy, University of Torino. Her research activities are mainly focused on the following topics: development of voltammetric and spectroscopic procedures for the determination and speciation of trace metals; study of element fractionation in sediment and soils from remote and anthropized areas; characterization of metal content in pharmaceutical formulations; interaction between trace metals and plants; application of chemometric techniques for the processing of experimental results related to the above mentioned matrices.

Mery Malandrino received her PhD in Chemical Sciences in 2001 at the University of Torino. She is presently Researcher of Analytical Chemistry at the Faculty of Sciences of the University of Torino. She is mainly involved in the development of sensitive analytical procedures capable of characterising complex environmental matrices, such as soils, sediments and atmospheric particulate matter. She uses chemometrics to interpret experimental results. The purpose of her studies is to gain insight into the behaviour of elements in uncontaminated ecosystems and their influence on climate changes. Furthermore, she studies the development of eco-compatible decontamination procedures for polluted soils.

∗ Corresponding author. Tel.: +39 011 6707844; fax: +39 011 6707615. E-mail address: ornella.abollino@unito.it (O. Abollino). 0003-2670/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.aca.2010.12.020

g. Reagents with different chemical properties are applied. The combination of the complexity of the matrices and phenomena under study with the generation of large data sets renders interpretation difficult. the remediation 1. meanwhile identifying their potential negative impact through their release into other environmental compartments and entry into the food chain. called ‘soft’ or ‘mild’ extractants. two reviews on element extraction from soils and sediments to date have included a chapter devoted to the application of chemometrics to the experimental results [1. This review describes the application of chemometric techniques in research studies involving single and sequential extraction treatments on soils or sediments. lacustrine ecosystems. is the multivariate technique most extensively used in conjunction with extraction assays. and cannot be assessed only from the values of the total concentrations.05 M ammonium EDTA. the visual illustration of the experimental results. the interaction between trace metals and clays. since they remain indispensable for a correct and complete data processing and interpretation. and many research studies also report the main properties of the considered matrices (such as pH. NaNO3 and BaCl2 . Measurements and Testing (SMT) Program (formerly BCR) developed and validated a single extraction protocol (0. despite concerns of being over-aggressive for this purpose (see next paragraph). Edoardo Mentasti has been full professor of Analytical Chemistry at the Faculty of Sciences of the University of Torino from 1980 to 2007 and is now retired.3]. the aims of the research studies in which they were used and the main findings obtained with their application will be addressed. Her research activities are finalized to study the behaviour of metals in different natural matrices (soils. Now she is holder of a research grant at the Department of Analytical Chemistry of the University of Torino. to develop new procedures of remediation of contaminated soils. This results in a process that is more time consuming than single extractions.O. but one that provides the partitioning of the total element contents into fractions of different availability [1–4]. Part I of the review will be focused on variable reduction methods for pattern recognition. to predict the plant-available fraction of elements: therefore the . Extraction assays allow us to obtain reliable and useful information only if the experiments are correctly planned and executed and if the results are properly interpreted. 1 h. the development of voltammetric methods of analysis. even when sophisticated multivariate techniques are subsequent applied. so that elements are leached according to different mechanisms. soil texture) which are important in order to understand their behaviour.1.unbuffered salts. They provide information on the mobility and availability of metals and other elements. sediments. Chemometric techniques can be a valuable tool in connection with single and sequential extraction procedures for several purposes: the development and optimization of the extraction conditions. usually in order of increasing strength. soils) with the aid of chemometric techniques. but no extensive treatment of this subject currently exists. sludge. models. the proportion amenable for plant uptake. similarities and differences among samples. acidification or complexation. / Analytica Chimica Acta 688 (2011) 104–121 105 Agnese Giacomino graduated in Chemistry in 2003 at the University of Torino. room temperature) [12].g. in which both advantages and disadvantages are ascertained. 2. especially when coupled to rapid multielement analytical techniques. which take into account the behaviour of multiple vari- ables simultaneously.01 M CaCl2 . according to their chemical properties. His main research interests are the development of preconcentration and speciation procedures for trace metal ions coupled to determination by atomic spectroscopy. content of organic matter. one of which. the acquisition of different areas of information. She received her PhD in Chemical Sciences in 2007 at the University of Torino. diluted acid or salt) is used to treat the sample and measurement is made on the amount of elements released from the matrix of interest [2. 3 h) [4]. of contaminated soils. To our knowledge. Extraction treatments give rise to large amounts of data. they are employed for the purpose of estimating plant-available fraction of elements [9–11]. such as soil. the applications of univariate and bivariate chemometric methods are reported. He has been Editor-in-chief of the Journal Annali di Chimica from 1996 to 2006. Following an overview on extraction procedures. They are regarded as more suitable than more aggressive extractants. the characterization of environmental compartments (seawaters.2]. such as chelating agents and acids. vegetables. CaCl2 . the calculation of element fractionation. to characterize the composition of pharmaceutical formulations. Mobility and availability depend on the reactivity and on the binding behaviour of elements with the components of the matrix. risk assessment and predictions of future events [5–7]. A single extracting reagent (normally a ligand. A more detailed overview on the properties and behaviour of the elements under investigation may be achieved through the utilization of sequential extractions. seawater). A SMT protocol exists (0. Introduction Single and sequential extraction procedures are widely used for the investigation of solid matrices. Overview on single and sequential extraction procedures 2. using chemometrics for data processing. in the case of soils. it should be emphasized that also univariate and bivariate chemometric methods are important. causes of the observed behaviour (e. e. sediments. nevertheless. fly ash and atmospheric particulate matter [1–4].ligands. source identification). namely principal component analysis (PCA). Chemometrics is deemed to be particularly advantageous when dealing with complex systems. then the principles of the multivariate chemometric techniques most frequently adopted. to study analytical methods for the determination of trace metals. . Abollino et al. We are confident that the present work will be of use to researchers interested in adopting the powerful tools of chemometrics in order to exploit the potentialities of single and sequential extractions of elements from solid matrices. such as soils and sediments. The Standards. Single extractions The main extracting reagents used in single extraction procedures can be classified. mainly diethylene triamine pentaacetic acid (DTPA) and ethylene diamine tetraacetic acid (EDTA). as: . due to the possibility of using multivariate techniques. Single extractions may be used for estimating the most potentially mobile element fraction and/or. sediment. such as CH3 COONH4 . In particular. including relationships among variables.

pH 2. 3.2 M CH3 COONH4 ). in crystalline Mn oxide or coprecipitated (0.1 M NH2 OH·HCl in 0. summing up of the first three steps. carbonate-bound (1 M CH3 COONa/CH3 COOH. bound to Fe–Mn oxides (0. [33–44. reagents for lithogenic fraction.2 M (NH4 )2 C2 O4 . in crystalline Fe oxide or coprecipitated (Na2 S2 O4 ). lithogenic (aqua regia/HF). 85 ◦ C. oxidable and sulphidic (H2 O2 twice. 1 M CH3 COONa/CH3 COOH.2 M CH3 COONH4 in 20% v/v HNO3 ). Ref. bound to carbonates (1 M CH3 COONa/CH3 COOH. pH 7).1 M NH2 H·HCl/HNO3 .2 M H2 C2 O4 . 1 M HNO3 ). 0.1 M NaOH). bound to carbonates (1 M CH3 COONa/CH3 COOH. elements nominally bound to carbonates oxides and reducible phases extracted in the same step. bound to carbonates and easily reducible phases (0. [94] [53] Sequential extraction scheme for Fe. Scheme 3: soluble.2 M CH3 COONH4 in 20% v/v HNO3 ). For these characteristics it is the most extensively applied sequential extraction technique nowadays. 5-step sequential extraction: exchangeable (1 M Mg(NO3 )2 ). HClO4 . oxidisable (8. 3.2 M CH3 COONH4 /HNO3 . mobilisable (0. associated to oxides of Fe and Mn (0. exchangeable and associated to carbonates (H2 O. chemisorbed on Fe. oxides and reducible (0.131] Developed by SMT in order to harmonize fractionation procedures and ensure comparability. 85 ◦ C. bound to Fe-oxides (0.1 M NaOH + sonication). associated to organic matter and sulfides (HNO3 /H2 O2 . 96 ◦ C).04 M NH2 OH·HCl in 25% v/v CH3 COOH/HNO3. oxidisable (HNO3 /H2 O2. except for the concentration of CH3 COONH4 . 2-step sequential extraction: mobile (0.6 M HCl. [45] Short procedure.1 M HNO3 /H2 O2 . Most of the other procedures derive from it. Mn and Al.1 M HNO3 ). It is distinctly different from procedures for metals. 4-step sequential extraction: soluble and exchangeable (H2 O). pH 2).2 M CH3 COONH4 in 20% v/v HNO3 ).116. pH 5). pH 2).106 O.04 M NH2 OH·HCl in 25% v/v CH3 COOH). bound to Fe–Mn oxides (0.11 M CH3 COOH).88. bound to acid-soluble residue (6 M HCl. Ca-bound (1 M HCl). 3. pH 7). 85 ◦ C). carbonates. Procedure Tessier’s sequential extraction: exchangeable (1 M MgCl2 . HCl). chemisorbed on humic acids (0. residual inorganic (HCl). exchangeable fraction following Kersten and Förstner’s scheme [32]. [50] Uncommon fractionation pattern. pH 2). The reagents and the operational definitions of the fractions substantially differ from the popular Tessier’s and BCR schemes. elements nominally bound to carbonates and easily reducible phases extracted in the same step. Sequential extraction scheme for P. pH 2). most reagents are different from Tessier’s scheme. bound to Mn oxides (0. pH 7). residual (HNO3 /H2 O).5 M NH4 OH·HCl. Based on Tessier’s scheme. residual (HClO4 /HF. 3. soluble and exchangeable elements extracted in the same step. pH 2. residual (HF/HClO4 ).49.2 M CH3 COONH4 in 20% v/v HNO3 ). bound to moderately reducible phases (0.01 M CaCl2 ). Scheme 1: exchangeable and soluble in water and acids (0.2 M (NH4 )2 C2 O4 /0. Al and organic moderately labile. pH 2). owing to the anionic nature of P compounds. It provides a detailed description of operative conditions. associated to organic matter and sulfides (HNO3 /H2 O2 . 1 M CH3 COONH4 . bound to organic matter (HNO3 /H2 O2 . recommended (aqua regia). Comment One of the first sequential extraction procedures developed. [46–48] Uncommon fractionation pattern.04 M NH2 OH·HCl in 25% v/v CH3 COOH.175 M Na-citrate/0. 4-step sequential extraction: exchangeable and acid-soluble (1 M CH3 COONa. reducible (0. [40] Scheme 2: based on Tessier’s scheme. 85 ◦ C. Modified Tessier’s sequential extraction: exchangeable (1 M CH3 COONH4 . pH 3). Revised BCR sequential extraction: exchangeable. pH 5).5 M CH3 COONH4 ).25 M NH2 OH·HCl. 96 ◦ C). pH 5). water. [52] Scheme 1: steps 1–3 according to the original BCR scheme. The original reference reports also the possibility to use 1 M CH3 COONa. absence of the residual fraction. pH 1. residual. within small stable aggregates. Proposed as an alternative to Tessier’s scheme for the measurement of the labile element portion. bound to organic matter.5).5 M NaHCO3 ). The differences are: distinction between Mn and Fe oxides. 3.54] Based on Tessier’s scheme. Abollino et al. pH 7. pH 5). 5-step sequential extraction: exchangeable (1 M CH3 COONH4 . [44. 2. 5-step sequential extraction: exchangeable (1 M CH3 COONH4 . following Kersten and Förstner’s scheme [32]. reducible (0.04 M NH2 OH·HCl in 25% v/v CH3 COOH. 85 ◦ C.1 M NH2 OH·HCl in 0. 0. The difference is the summing up of the first two fractions. oxidisable (H2 O2 . / Analytica Chimica Acta 688 (2011) 104–121 Table 1 Sequential extraction procedures adopted in the papers cited in this review.136] [49.3 M CH3 COONH4 ).2 and 0. 1 M CH3 COONH4 . pH 5). H2 O/H2 C2 O4 ). 3. Scheme 2: exchangeable and associated to carbonates (1 M MgCl2 . . inorganic moderately labile.01 M HCl. residual (HClO4 /HNO3 ).and acid-soluble (0. 1 M CH3 COONa. pH 7). HNO3 ). bound to organic matter and sulphides (HNO3 /H2 O2 . residual (HF/HCl/HNO3 . cooling. pH 2. The difference is in the exchangeable fraction. residual (HNO3 /HCl/HF/HClO4 /H2 O2 ).11 M CH3 COOH). organically bound or associated with organic matter (0.80. reducible (0.8 M H2 O2 /HNO3. pH 2. exchangeable fraction following Kersten and Förstner’s scheme [32]. 85 ◦ C. 96 ◦ C).025 M H-citrate for the first and third fractions respectively. The differences are: exchangeable fraction following Kersten and Förstner’s scheme [32]. pH 8. 6-step sequential extraction labile inorganic and labile organic (0.95. pH 2.005 M DTPA).3 M Na2 S2 O4 /0. 3. bound to organic matter (0. pH 2. pH 4. Based on BCR scheme. pH 2. pH 5). pH 2. The differences are: presence of a step for water-soluble elements.7 M NaOCl). bound to organic matter (HNO3 /H2 O2 . residual organic (HCl/H2 O2 ).04 M NH2 OH·HCl in 25% v/v CH3 COOH.130. It was the most extensively applied scheme before the introduction of the BCR protocol. physically inaccessible and within small stable aggregates.2 M CH3 COONH4 ). Scheme 3: based on Tessiers’s scheme. associated to oxides of Fe and Mn (0. physically inaccessible (0.04 M NH2 OH·HCl in 25% v/v CH3 COOH).

/ Analytica Chimica Acta 688 (2011) 104–121 Table 1 (Continued) Procedure 4-step sequential extraction: loosely sorbed (1 M NH4 Cl). with the purpose of harmonizing the quantification of the extractable trace-metal contents in soils and sediments [3].69] Aimed at studying element bioaccessibility.5. 0.2). an additional first stage can be added to Tessier’s scheme so that it would be possible to measure the water-soluble element fraction [27].1 M NaOH). simulated small intestine conditions (pH 7 with NaHCO3 . which are summarized and briefly commented in Table 1 together with the other procedures adopted in the papers cited in Sections 4 and 5. indeed. The BCR scheme has shown steady increase over time due to its advantages over other current sequential protocols. 0. Acids at high concentrations (e. weakly sorbed-bioavailable organic and inorganic (0. 4-step sequential extraction: soil solution and labile (water and anion exchange resin). and enables comparability among data obtained in different laboratories with different samples. Some researchers measured element mobilization by the Toxicity Characteristic Leaching Procedure (TCLP). Simplified PBET (SBET): 30. pH 2. Sequential extraction scheme for P. readily soluble nonexchangeable (0. addition of 175 mg bile salts and 50 mg pancreatine to the simulated stomach fluid).2.diluted mineral acids. the latter are secreted as metabolic products through plant roots. Sequential extraction scheme for K. Notably. Aimed at studying element bioaccessibility. inorganic. acid-soluble (0. in order to permit calculation of the recovery by comparison between the sum of the amounts extracted into the four fractions and the pseudo total content obtained by aqua regia digestion [1. pH 8. water and acid-soluble. which involves the use of HClO4 and HF. a fourth step of digestion with aqua regia is recommended. most of them bear close resemblance to BCR and Tessier’s protocols and.03 g L−1 glycine/HCl. small intestine 1 (pH 7 with NaHCO3 . Comment Sequential extraction scheme for P. 150 ◦ C). The reagents and the operational definitions of the fractions substantially differ from the popular Tessier’s and BCR schemes. so it was revised by changing the concentrations of some reagents and some operative conditions.O. have been fashioned on these procedures. in humic and fulvic acids and in Al and Fe phosphates (0. The reagents are similar to the ones reported in the previously mentioned schemes.25 g pepsin. or low molecular weight organic acids. reducible. therefore enabling a uniform approach in different laboratories. 37 ◦ C. It simulates gastrointestinal tract environment.2 M sodium tetraphenylborate). pH 1. associated to humic matter (0. residual [23]. metal oxide bound (1 M NaOH). It is mainly focused on the exchangeability of the element. water-soluble (H2 O).g. calcium bound (0.1 M NaOH). bound to oxides of iron and manganese.11 M NaHCO3 /Na2 S2 O4 ). but the definitions of the fractions are different. but this procedure is not so common. strongly bound chemisorbed-potentially bioavailable (0. [55] [118] [66.2 M NaOH). the scheme takes less time and is simpler than Tessier’s procedure. 2. small intestine 2 (the same as small intestine 1 after an additional 2 h incubation). Abollino et al. 0. Sequential extractions The most popular sequential extraction procedures are Tessier’s and BCR schemes. apatite or Ca-bound and non-bioavailable (1 M HCl). [67] use of unbuffered salts has notably increased over the last 10 years [13–16].5 M NaHCO3 . As with the other sequential extraction schemes. this happens when it is determined by means of strong acid extraction (e. Ref. Characterized by the use of an anion exchange resin in addition to chemicals.5.17. the fifth consists of elements with low mobility and which are unlikely to be solubilized under natural conditions over a reasonable period of time. 420 ␮L lactic acid and 500 ␮L acetic acid made up to 1 L with H2 O/HCl. 37 ◦ C). 3-step sequential extraction exchangeable (1 M CH3 COONH4 ). Tessier’s protocol provides the partitioning of elements into five operationally defined fractions: exchangeable.24–26]. hardly soluble (1 M H2 SO4 ). the US EPA’s method for testing waste toxicity [20].2).5. such as malic and citric acids.g. addition of bile salts and pancreatine to the simulated stomach fluid). 0. bound to organics and sulphides. Three fractions are obtained with the BCR scheme: exchangeable. 6-step sequential extraction: organomercury (CHCl3 . A number of procedures differ mainly in the reagent used to .5 M HCl). [56] 107 [57] Sequential extraction scheme for P.25 g pepsin. hence they are believed to simulate natural conditions [17–19].g. organic and microbic (0. oxidisable. Modified PBET: stomach phase (simulated stomach fluid: 1. even if it is not officially part of the protocol. This is thanks to a method that includes a detailed and highly methodical procedure for the preparation of the reagents and implementation of the extraction process. reductant soluble (0. It differs from the above cited PBET method for the presence of two steps in simulated intestine conditions. The first version of the protocol showed a low reproducibility in interlaboratory exercises.05 M HCl. Physiologically Based Extraction Test (PBET): simulated stomach conditions (simulated stomach fluid: 1. aqua regia) and not after total mineralization as in Tessier’s scheme. It differs from the scheme for P cited above for the nature and/or concentrations of the reagents and for the definitions of the fractions. elemental (aqua regia. bound to carbonates and specifically adsorbed. and the availability of a standard reference material for the validation of the results. recalcitrant nonexchangeble (0.01 M HCl). Whereas the first fraction is quite labile and therefore easily available for plant uptake.5 M NaHCO3 . It considers only one gastric phase. residual. [58] Sequential extraction scheme for Hg. pH 8.50 g sodium citrate.50 g sodium malate. in the revised version. It should be pointed out that in many of the procedures reported in Table 1 the so-called “residual” fraction is actually a “pseudo-residual” fraction. 420 ␮L lactic acid and 500 ␮L acetic acid made up to 1 L with H2 O/HCl. Many other sequential extraction procedures were developed. [68] Aimed at studying element bioaccessibility. 6 M HCl) were also used to evaluate the mobile portion of elements [22]. 0. The BCR protocol was developed by the SMT Program within the framework of a collaborative project.01 M Na2 S2 O3 ). 37 ◦ C). 4-step sequential extraction: plant-available and water-extractable (H2 O). . pH 2.50 g sodium malate. 0.50 g sodium citrate.5 M HCl). a decrease of element availability during extraction sequence is presumed. labile. which involves a single extraction with diluted acetic acid and sodium hydroxide [21]. HgS (aqua regia). e. Notably.

leach*. the concentrations of the analytes. it must also be said that valuable information on the behaviour and mobility of elements can be achieved. Different approaches can be distinguished: (i) in most cases. a sample.56–59]. in order to cover all the types of analytes considered in the studies described. 4. 3. lake sediment [36. 0. or “potentially toxic elements”.. such as “metals”. but its high sensitivity allows the dilution of the sample solutions before analysis. or major metals such as Fe. food chain and consequently human health. Of course. BCR CRM 701. General considerations 3. in order to describe the properties of the investigated system or for risk assessment. i. Analytical aspects The methods of determination of the elements present in the extracts are well established. soil*. Most extraction procedures are designed for cations and are not completely suitable for As. their association with matrix components is often questionable [1. etc. Table 2 shows that most authors used atomic spectroscopic techniques. In most chemometric techniques the data are arranged in matrices: each row corresponds to an “object”. the elements determined. simulating the conditions of digestion in the stomach and in the small intestine respectively.41]. mainly ICP-AES.). in particular the nonselectivity of reagents and the occurrence of readsorption and redistribution phenomena along the extraction sequence [77]. they are less commonly applied to single and sequential extraction results. Ca and Mg. multivar*. “trace elements”. 1 M KNO3 . which in many cases reproduce Tessier’s protocol but differ in the first steps [40. hence the studies using two or more of such techniques will be mentioned two or more times. sedim*. texture. as a confirmation of its popularity [33–44]. mobil*. We fully support Bacon and Davidson’s conclusions that sequential extractions will have a healthy future in the 21st century. and enable the on-line coupling of the extraction and determination stages. Some examples of chemometric treatment of bioaccessibility data will be cited in Part II of the review [64–69]. 12 M HCl and 4 M HNO3 ) [27–31]. Some procedures were designed for the fractionation of single elements. Finally. 0. and would be available for absorption [60–63]. putting aside the schemes’ limitations. being focused on the testing of a chemometric technique. when some of them are investigated in the cited papers.g. like cation exchange capacity (CEC).1.3. and reports the values of the “variables”. Terminology We will use the term “elements”. Furthermore. Other studies were based on different schemes. 1 M Mg(NO3 )2 . This is because they provide a better simulation of natural conditions and give information on the kinetics of element mobilization.1. Presently. such as gastrointestinal content or perspiration. We . in particular by analyte readsorption. that is the fraction of a chemical that is liberated in relevant biological fluids. the software package used. fraction*. sequential extraction methods can provide a valuable tool to distinguish among trace element fractions of different solubility related to mineralogical phases” [78]. which was common until about 10 years ago. As the extracted fractions are operationally defined.76].32].80] or. However. In addition. the chemometric techniques described in this work are suitable for application to the results of dynamic procedures. whereas chemometric techniques are extensively used to process data on total element concentrations [e. mentioned by Bacon and Davidson in their review “Is there a future for sequential chemical extraction?” [1] affirms: “despite some drawbacks.g. pH. speciat*. Table 2 summarizes the main features of the papers considered in Sections 4.4]. including nonmetals such as As and Se. the analytical method used for the determination of the extracted elements. instead of more specific words. other methods attempt to distinguish between elements bound to Mn and to Fe oxides. and the results are influenced by the experimental conditions adopted (method of sample drying. despite having a lower sensitivity than the latter. the procedures (especially in the batch mode) are time-consuming. in addition. / Analytica Chimica Acta 688 (2011) 104–121 evaluate the exchangeable fraction (e. total nitrogen. shaking device. but that their results will be useful only if they are interpreted with full awareness of their limitations [79].e. which has the advantage of being more rapid than FAAS and GFAAS. Chemometric aspects We covered the literature from 2000 to 2009. We will refer to “main properties” to indicate one or more physico-chemical characteristics of soils or sediments.g. organic matter.108 O. ICP-MS was used in some researches. [8. i. and use a data set mainly as a means 3. It is important to note that sequential extraction procedures have several drawbacks.69–74].g. the use of flow-through dynamic approaches has been increasing over the last decade [84–86]. electrical conductivity. chemometrics is used as a tool for the interpretation of the experimental results.87–90]]. Most of the studies cited in this work used the conventional batch procedure for the extractions. but is nowadays referred to the determination of well-defined chemical species. Quality control of the experimental results for sequential extraction is assessed by comparison between the total content and the sum of the extracted fractions [51. the extraction media.44–54]. in that sample. Other forms of sequential extraction yield the so-called bioaccessibility. such as Hg [55] or P [53. but some examples of column leaching are also reported [81–83]. 3. Some sequential extraction methods for As exist. will not give any judgement on the suitability of the extracting solutions used.38. We will follow recommendations by the IUPAC and refer to the results of single and sequential extractions as “distribution” or “fractionation” of elements. They usually involve two steps.38.e. An IUPAC report. The BCR protocol was used in many of the studies on sequential extractions cited in this review. the high concentrations of dissolved solids present in many extracts may be a drawback with this technique.01 M NaNO3 ) or to extract elements bound to organic matter (e. In fact. which will be described in Part II [66. using different concentrations of reducing agents or different temperatures [28–30. or on the correspondence between the definition of the extracted fractions and the actual content of the extracts.1 M K4 P2 O7 or a mixture of KClO3 . as our focus is the data treatment techniques. A different approach to sequential extractions is the use of non-specific reagents coupled to chemometric data treatment. or “features”. (ii) some papers take a counter approach. searching through ISI Web of Knowledge with combinations of the following keywords: extract*. chemometri*.2. e. they are seemingly less affected by the drawbacks of batch extraction procedures. which are based on its similarity with P [75. organometallics or metals with different oxidation states [3]. by using a standard reference material certified for the extractable content. for the BCR scheme. owing to its presence in anionic form in soils and sediments. Abollino et al. the chemometric technique(s) applied.2 and 5 of the review: the type and location of the investigated samples. avoiding the term “speciation”. In this review (Part I and Part II) the various chemometric techniques will be discussed one by one. and such procedures aid in establishing potential risks to the environment.

0. Ca.01 M CaCl/0. Soil from an experimental area (Volperino. Sr. 0.0 [115] Correlation analysis. Fe. Ni.1 M TEA pH 7. 0. Greece) Lake sediments (Volvi and Koronia Lakes.015 M NH4 F. Ni. Mg.0S [56] Contaminated soil treated with organic residues (Aljustrel mining area.0 [22] Ca. [58] Al.005 M DTPA + 0.1 M Ca(NO3 )2 . Zn FAAS.1 M HCl) Single extraction (0. ICP-MS Correlation analysis. Zn FAAS Single extraction (0.129] O. 0. HCA Statistica 6.2 M CH3 COOH. Pb. MLR SPSS 11. Greece) Elements P Analytical technique UV–vis spectro-photometry Extraction procedure 4-step sequential extraction. Co. Cu. Mg. P: 4-step sequential extraction Single extractions (0.02 M EDTA + 0. Co. Cu. Mn. 0. HCA.65. PCA Canoco 4. GF-AAS ANOVA.3. Ba. PCA SPSS 8. Ce. Pb. 0. The papers are arranged in the order in which they were cited in Sections 4 and 5. Cd. Ni Pb. speciation within each extract Single extraction (Al. 0. 1 M CH3 COONH4 ). Ni. Mn. Spain) Cr. U. 0. Nb.01 M CaCl2 pH 5. Korea) ˜ Spain) Forest soils (La Coruna. Rb. / Analytica Chimica Acta 688 (2011) 104–121 Cd. W. Zn FAAS ANOVA. Fe. Zn Fe. 1 M NH4 NO3 . geostatistics GS+ [102] Soils from a national database project (Ireland) Roadside sediments (Seoul. Th. Cr. K.5 M CH3 COOH. Mg. Cu. Pb. Italy) K. single extractions (0. Fe. Mn ICP-AES. Zn Co. Mn. Cr.01 M CaCl2 ) Single extraction (6 M HCl) Single extraction (Mehlich III extractant: 0. 0. Mn. FA Correlation analysis. PCA. As. V. Ge. Surfer Not reported [111] [114] [113] Correlation analysis. Abollino et al. PCA. Winn Parish.5 M HNO3 . geostatistics.25 M NH4 NO3 . UV–vis spectro-photometry ANOVA. Hg. Zn FAAS Correlation analysis.05 M EDTA) Single extraction (Fe and Mn: 0. geostatistics Correlation analysis.5 M CH3 COO NH4 . P FAAS.0 [37] Contaminated soils (along Dommel River. Ga. Se. Mo. FKA Not reported KRIGE. Matrix Surface river sediments (Louros River. Fe. Fe. Cr. GF-AAS Urban and olive oil sludge (Ourense town and province of Jaén.01 M CaCl2 + 0. GF-AAS. Mn ICP-AES Correlation analysis. Mg. 0. Cd. Tl. Ti. Sc. COKRI. Fe.1 M NaNO3 .013 M HNO3 . Correlation analysis.001 M EDTA) Single extraction (acetate buffer) Chemometric treatment ANOVA. 0. Sn. Pb.5 M CH3 COONH4 + 0. Cu. 0. Portugal) Cu. Li. PCA SPSS 10. Fe. 0.11 M CH3 COOH.5 M CH3 COOH pH 4. HCA. BCR. S. 1 M CH3 CONH4 . Ca.0 Ref.005 M DTPA/0. La. Mediterranean Sea) Forest soils (Kitsatchie National Forest. Y. BCR Single extraction (0. Belgium) Cd ICP-AES. PCA Statistica [39] 109 . 1 M MgCl2 . Mg. Na. LDA Software SPSS 13. HG-AAS ICP-MS ICP-AES Soils with different composition and pollution levels (Flanders. 0. Mn. Zn) As. the Netherlands) Muddy and sandy marine sediments (Egyptian coast. Cr. Ca.5 [112. P (plus total Al. aqua regia) Single microwave extractions (with BCR extractants). Louisiana) Cd.1 M HCl. HCA SPSS 14 [107] ICP-AES.Table 2 Selected applications of chemometric techniques to single or sequential extraction results. Sb.02 M EDTA).7. Ta.1 M tetraethylammonium Soil solution. Correlation analysis. Ni.01 M CaCl2 .

geostatistics Variowin 2.43 M HNO3 . Co. 0. Na. pH 7) Single extractions (0. Pb. Cu. Belgium) Contaminated soils from an abandoned mining area (Salsigne. France) Roadside soil and dust (Sweden) Agricultural soils irrigated with wastewater (Hidalgo State. Cu. Chishui. Zn ICP-AES.05 M EDTA. geostatistics Correlation analysis.110 Table 2 (Continued ) Matrix River sediments (Ell-Ren River. Abollino et al. Pb. MLR VARIOWIN 2.2 [117] Soils from rice fields (Takatsuki City. Ni. Zn. Cu. Mg. Southern Taiwan) River sediments (Haihe River. Fe. Xl-Stat 5. Zn ICP-AES AAS Correlation analysis. Ca. Cu. Cr. As.1 M Ca-acetate/0. PCA. Peikang rivers. Excel 9. FES Correlation analysis. PCA.Mo. U.0 for Windows Not reported Ref. Pb ICP-MS 4-step sequential extraction Extraction with water followed by Tessier 6-step sequential extraction for Hg Correlation analysis. Zn ICP-AES. Ni. Pb. ICP-MS BCR Correlation analysis. Cu.0 [104] [110] [120] P Colorimetry Correlation analysis.05 M EDTA). Zn Mn. Zn B. Pb. China) River sediments (Yenshui.04 [119] As. Fe. Ni. Ca. Ni.0. Fe. 0. PCA Correlation analysis. Zn Cd. HCA VESPER 1. Cr. geostatistics Correlation analysis. Pb. 0.0. ANN Statistica 6. P FES.5 M CH3 COONH4 /0. Ni. Mexico) River sediments and floodplain soil (Warta River. Mn. Ni. Spain) Agricultural cambisol (Kassow. HCA SPSS for Windows 10 [52] Cr.2 Statgraphic Plus 5. Cd.01 M CaCl2 . colorimetry Correlation analysis.3 M CH3 COOH) 3-step sequential extraction (K). Fe. Cd. Cr. (Fuyang County. Cu. Japan) K. Cu. modified Tessier. aqua regia. Co. FAAS 5-step sequential extraction Tessier Single extraction (0. Co. China) Agricultural soils with vegetable crops (Lower Vinalopò region. PCA SPSS 10 [51] Hg CV-AFS Correlation analysis. Cd.2 [35] Al. Zn Cu Cd. CaCl2 . K. N.UK) Uncontaminated soils (Central Spain) Soil affected by the Aznancóllar mine spill (Spain) Surface soils from seven land uses. Cr. Mn. PCA The Unscrambler 7. Na. [48] [116] [47] Cu. Cd. / Analytica Chimica Acta 688 (2011) 104–121 As. Mg. single extractions (N: 2 M KCl) Soil solution Correlation analysis. Mn. HCA SPSS 10. K.5 M CH3 COOH/0. Ni. Panˇ cevo Oil Refinery. Germany) Elements Cd. Cr. Co. Ba. Mn.0 [95] [44] O. W.0 SPSS 13. K. PCA Software Not reported SPSS 12. BCR.0. Tsengwen. Ca. Pb Co. Zn. PCA Correlation analysis. Pb. Surfer 6. Cd. Potzu. water and Bray method (P). Mn. Cr. Analytical technique Not reported GF-AAS AAS Extraction procedure 5-step-sequential extraction Tessier 5-step sequential extraction Chemometric treatment Correlation analysis.1 M Ca-lactate/0.6 SPSS 10. Serbia) Soil from a landfill (Bedfordshire.01 [50] ICP-AES Correlation analysis. MLR SPSS 10. southern Taiwan) Alluvial river sediments (Danube river.02 M Na2 -EDTA. Single extraction (0. Mg.0 [55] . Poland) Al.05 M EDTA) Single extraction (DTPA. TEA) Single extraction (0. Cu. Ni. Cu. Rb. geostatistics GS+ [118] Contaminated and uncontaminated soils and sediments (Flanders. Pb. Pb. Zn Not reported AAS FAAS. geostatistics Correlation analysis. V. Sr. GF-AAS Correlation analysis.

Cu.5 GSLIB WLSFIT. Pb. Pb. Turkey) and grapes Ca. N. P FAAS FAAS. GIS Not reported + [105] [108] O. [57] Cd. Pb. S. P:NaHCO3 ) Single extraction (Al.1 M Na2 EDTA. K. Mg. HCA Correlation analysis. P: 2. VARIOWIN 2. Cr. Fe. Mg. Mg. FAES AAS ICP-AES. Cu. aqua regia) Single extraction (0. Fe. Fe.21 StatView for Apple Macintosh Not reported [127] [34] [46] Cd. Mg. double lactate extraction (K.05 R + [122] [124] [33] [100] Cd GF-AAS Not reported [126] Soils from around a zinc smelter (Kayseri.05 M Na4 P2 O7 ) BCR Modified Tessier’s procedure Chemometric treatment Correlation analysis. MLR Statistica 6-0 Canoco 4. Mn. Pb. Co. Spain) River sediments (Yenshui.02 M AAAC-EDTA.5% CH3 COOH) Single extractions (0. Cu. P Not reported Colorimetry FAAS Colorimetry GS+ GS . geostatistics Correlation analysis. ArcGIS SPSS 9. Zn N. geostatistics Correlation analysis. geostatistics Software Surfer Ref. P Analytical technique ICP-AES Extraction procedure 4-step sequential extraction (P). N. RDA Correlation analysis. Ni. Cu. Galicia. Mn. Peikang rivers. Abollino et al. Na. Cr. geostatistics Correlation analysis. HCA PCA XLSTAT BMDP [38] [49] 111 . PCA. P K. Fe: oxalate) Single extractions (K: 1 M CH3 COONH4 ) BCR Single extraction (N: 2 M KCl. Zn FAAS Correlation analysis. geostatistics Correlation analysis. / Analytica Chimica Acta 688 (2011) 104–121 [123] [121] Al. 0. PCA.5 M KCl. Cu. geostatistics. Cu. Zn ICP-AES.025 M H2 SO4 . Spain) Al Cd. Cr.005 M DTPA-TEA. FKA Correlation analysis. Pb Ca. Na. Ca. geostatistics PCA PCA GS+ . southern Taiwan) Marine sediments (Terra Nova Bay. Mn. expressed as SEM. Si (+ anions) Zn Al. Zn N. 0. Fe. P) Column leaching with modeled acid rain Column leaching with water DTPA Single extraction (1 M HCl. ICP-MS Correlation analysis. Ni. Ni. P K. P. geostatistics Correlation analysis. geostatistics Correlation analysis. Mn. Fe. Zn Cd. HCA SPSS 10. Ni. geostatistics Correlation analysis. Na:CH3 COONH4 . Pb. Potzu. Bohemia. 0. northeastern Germany) Soil profiles (Torun. P Cd. Chishui. Cu. Pb Co. Cd. Mn. see text) Single extraction (1 M) Single extractions (Cd: DTPA) Single extractions (K: 1 M CH3 COONH4 ) Single extractions (K. Zn ICP-AES GF-AAS Not reported Correlation analysis. Pb. FAAS BCR Modified Tessier’s scheme PCA. Cr. P Cd. 1 M NH4 Cl Single extractions (0. Czeck Republic) Surficial river sediments (Louro River. Fe. Tsengwen. Surfer [82] [83] [91] [98] Colorimetry GF-AAS Colorimetry Not reported R 1.01 and 0. GF-AAS DPASV. Cr. Co. 1 M NH4 NO3 . 1 M CH3 COONH4.1 M HCl in 0. ANN Correlation analysis.05 M CaCl2 .8 for Linux GIS GS . Turkey) Forest soils (northwestern Spain) Agricultural soils (South Australia) Elements K. Ni. Poland) Calcaric Fluvisols (Spain) 587 soils (North Dakota) Lake sediments (The Netherlands) Forest soil (northwestern Spain) Agricultural soils (central Greece) Agricultural soils (northeast China) Agricultural soils from an experimental area (Italy) Forest soil (North Carolina) Agricultural soils (northeast China) Soils from urban garden (Kayseri. Antartica) Estuarine sediments affected by the Aznancóllar mine spill (Guadiamar and Guadalquivir rivers. Ni. N K.0 [125] Forest soils (Jizera Mountains.Table 2 (Continued ) Matrix Agricultural soils (Harz mountains. Ni. Cd. Cr. geostatistics Correlation analysis. 0.

Ca. Ni. Y.02 M EDTA in 0. Co. Pb. Cr. Ultisols.05 M EDTA. FA. Cr. PARAFAC. HCA Correlation analysis. Spain) Mine-impacted river sediments (Cacalotenango and Taxco rivers. Mg. Cu. Fe. India) Al As. GF-AAS ICP-AES BCR Scheme 1. Ni. Matlab 6 Matlab 7.0. Ni. Matlab 6 [97] [41] Al. Cr. FES: flame emission spectroscopy.0. Zn. Torino. TUCKER3 N-way methods Statistica 7. Mn. aqua regia). Ti. Fe. Cu. Mg. Pb. LDA: linear discriminant analysis. Zn Ca. Scheme 2. ICP-AES: inductively coupled plasma atomic emission spectroscopy. Cu. Fe. Mn. V. La. Al. HCA PCA. Germany) Soils irrigated with wastewater (Jajmau area. La. single extractions (Al: KCl. Sc. Pb. Fe. Zn Cd. Cr. Zn Cr. . Ca and Mg: CH3 COONH4 ) 1 M HCl BCR Chemometric treatment PCA. Scheme 3 BCR MA. Pb. Pb. Spain) Marine sediments (Mejillones del Sur bay. FKA: factorial kriging analysis. ICP-MS Tucker3 N-way method Not reported [42] AAS Single extractions (water. 0. Se. Zn ICP-AES. Cu. K. Fe. Cu. Buenos Aires.1 [94] [36] O. Ca. Mg. GF-AAS Correlation analysis. DPAdCSV: differential pulse adsorptive cathodic stripping voltammetry. Cr. HCA XLSTAT [88] ICP-AES ICP-AES. Argentina) Contaminated soil profiles (Bad Liebenstein.5.112 Table 2 (Continued ) Matrix Mangrove sediments (Mengkabong Lagoon. XRF: X-ray fluorescence.0. Co. Ni. P. GF-AAS PCA. Cu. N-way method PARAFAC N-way method Minitab 13. Eu. GF-AAS: graphite furnace atomic absorption spectroscopy. 0. Na. Cr. Austria) Soil collected around a coal-fuelled power plant (Velilla del Río Carrión. Mn. Ca. PARAFAC. Pb. Mo. Venezuela) Ferrite precipitated from polluted effluents River sediments (Pisuerga and Carrión rivers. Sb. Mn. 0. Zn ICP-AES. Cr. Mn. Mg: 1. Zn Pb ICP-AES ICP-AES DPAdCSV. Pb. Cu. Y. BCR 4-step sequential extraction BCR Tessier Modified Tessier extraction Column leaching with natural and acidified rainwater P: 11-step fractionation. FAAS: flame atomic absorption spectroscopy. Cd.5 M NaHCO3 ) 2-way PCA + unfolding. DPASV Not reported PCA PCA. Ni.5 M CH3 COONH4 ) 2-step sequential extraction Single extractions (0. K. PCA FA Multi-Variate Statistical Package 3. UK) Dust in schools (Caracas. Ni. Glasglow. HCA XLSTAT XLSTAT [80] [89] Superficial soil and grass (Gipuzkoa. South Africa) Alfisols. V. Ni. Argentina) Cd. Cu. Ni. Zn Al. Mn. Spain. Mn. GF-AAS PCA. HG-AAS: hydride generation atomic absorption spectroscopy. Cu. Cu. V. MA. Cu. [129] Al. Cu. Ni. Cd. Sc. other elements: aqua regia) Single extraction (0. PARAFAC. Cu. Mg: CH3 COONH4 . Fe. Mn. Mn: 5-step sequential extraction. Zr ICP-AES. Ti. Pb. Co. ICP-AES FA Statistica [81] Not reported FA. P: 0. Ti. Abollino et al. Zn Cd. Mn. Cd. ANN: artificial neural network. Mn.0 M NH4 Cl. Malaysia) Agricultural soils (Piedmont. DGT: diffusive gradients in thin films. Zn Al. Ni. Ni. Cr. Pb. Sabah. Co. K. V. Italy. Na. Zn AAS ICP-AES. Mexico) Gold tailing dumps materials (Witwaterands. Cu. Zn. IC: ion chromatography. DPASV: differential pulse anodic stripping voltammetry.0 [133] List of abbreviations: AAS: atomic absorption spectroscopy (the atomizer was not indicated). Italy) Contaminated soils (Piedmont. Ca. Ni. Cr. As. MLR PCA. TUCKER3 N-way methods Not reported Minitab 13. P XRF. Cr. Fe.5 M CH3 COOH. RDA: redundancy analysis. CV-AFS cold vapour atomic fluorescence spectroscopy. Tessier Tessier Single extractions (water. Oxisols soils (Misiones Province. Al. Cr. Italy) Contaminated soils (Piedmont. ICP-MS: inductively coupled plasma mass spectrometry. Cd. Zn Cd. Fe. Urban soils (Sevilla.5 M CH3 COONH4 ). MLR: multiple linear regression. Na. Fe. Cu.5 M HCl. Fe. Zn GF-AAS Not reported PCA PCA Statistica SPSS 11. Fe.1 Minitab 10X Not reported Not reported [45] [130] [131] [54] Al. Thuringia.0 [136] [40] Cd. Co.02 M EDTA in 0. Pb. / Analytica Chimica Acta 688 (2011) 104–121 Cd. HCA Software Not reported Ref. Ca. Mg. Ni. Pb. Cr. Matlab 7. Pb. Italy) Elements Al. Y. Spain). Zn Analytical technique FAAS Extraction procedure Single extraction (Na. K. LDA Not reported [53] Acidic forest soils (Tyrol. Pb. Mn. Mn. Chile) Estuarine sedimenst (Bahìa Blanca.

i.O. A similar application of non-parametric ANOVA was reported by Kaiserli et al. A different approach to data treatment was used by Alvarenga et al. 4. subtracting the column means from each value and dividing by the standard deviation [5]. for instance for its toxicity. presence of traffic. log 10 transformation is usually carried out in case of deviation from normal distribution.42. ANOVA ANOVA tests for the presence of systematic differences between groups of data differing for the value (“level”) of one or more parameter (“factor”). 4. using the Kolgomorov–Smirnov test in order to check for normal distribution [37.36. such as the percentage of variance explained by principal components (see Section 5. while other results. [58] used nonparametric ANOVA (Kruskal–Wallis ANOVA) in order to study the seasonal trend for P in fractions extracted from river sediments by testing for significant differences in sampling dates. x2 ) = and ¯ 1 )(x2 −x ¯2) (x1 −x n−1 cov(x1 .43. for instance about total concentrations. such as leaching by rain and vehicular traffic.58. we underline that such details are important. (iv) in a few cases. contamination. named “grand mean”) is split into two contributions.) or other situations (sampling time. [111] used two-way ANOVA to study the spatial (sampling site) and temporal (sampling month) variations in element concentrations in roadside sediments.53. [37].111]. Abollino et al.95. two-way or multi-way ANOVA are used. Univariate and bivariate techniques Chemometrics comprises not only multivariate techniques.55. who demonstrated that the sampling month had a significant effect on P fractions in lake sediments. Yun et al.99–101. The total variance of the data (calculated as the sum of squares of the deviation of the data from the total mean. the calculation of concentration means and standard deviations is obviously a prerequisite of any discussion and interpretation of data. normal distribution: such requirements can be found in textbooks and handbooks on chemometrics. laboratory.1. The examples reported are taken from the papers cited in Section 5.69–74]. x2 ) = where cov(x1 . When data do not follow normal distribution. i. The results of ANOVA were combined with those of the median test. they applied a post hoc Tukey honest significant difference test to further elucidate such differences. n = number of data (standard deviation) sx = n−1 In the presence of data not normally distributed. 107. the transformation of data [35. in terms of significant differences between sites. who applied parametric or non-parametric ANOVA depending on the results of Kolgomorov–Smirnov test for homogeneity of variance and normality. so we will treat this topic in Part I of the review.66. pH. Only the findings obtained with chemometrics on extraction assays will be reported. An example of extensive use of ANOVA is the paper by Bleeker and van Gestel [112]. i.92. First of all.g.e.49–51. the identification and elimination of outliers.106. then (Part II) we will describe the applications of chemometrics to the characterization and optimization of the extractions and to the calculation of element partitioning.83. If a single factor is being investigated.93].1. (iii) other studies use chemometrics in order to calculate the partitioning of elements among different components of soils or sediments [66. The strength of the association is usually expressed with Pearson’s correlation coefficient (r): r (x1 .2.94. ANOVA is a very important tool to describe differences between different extraction steps.91. such as the replacement of missing values. but also bivariate and univariate statistical methods. Preliminary statistical tests are sometimes carried out before applying multivariate data processing.94–110].81. They did not give indications on how ANOVA was carried on. within-groups and between-groups variance. Correlation analysis Correlation analysis is a bivariate technique for the measurement of the degree of association between two variables [5]. such as the presence of spatial or temporal variations. In the presence of two or more factors. sequential extraction procedures must sometimes be checked regarding one analyte because it is of particular importance. / Analytica Chimica Acta 688 (2011) 104–121 113 to demonstrate the efficiency of such technique [8. It should be stressed that most of such techniques can be applied only if the data fulfil some requirements. it can be used to examine environmental properties giving rise to differences in a dataset. Analysis of variance (ANOVA) and correlation analysis are treated in more detail hereafter owing to their valuable contribution to the interpretation of experimental results.91]. The combination of these methods was shown to provide a more representative description of seasonal patterns in complex data sets than discriminant analysis (see Part II). within a study.2): the latter permitted to identify the phenomena causing the variations identified with the former. throughout their study on the effect of variations in metal availability on earthworms. the nonparametric Spearman rank correlation coefficient can be used: r =1− 6 2 di ¯ )2 (xi −x n(n2 − 1) (2) where n = number of paired observations and di = difference between the ranks given separately to the two variables. will be omitted. . A remarkable aspect of their study is the coupling of ANOVA with factor analysis (see Section 5. who referred to the results of this technique.109. which find wide and crucial applications to the processing of results for extraction assays. or between samples or elements. Most of the studies cited in this paper use chemometrics for data visualization and interpretation.70.e. x2 ) sx1 sx2 (1) (covariance).2. for a correct interpretation of the results and assessment of their validity.g. Such contributions are compared with an F-test: if a significant difference is found. data are often scaled by column-standardization.e. then it can be concluded that the factor has a significant effect on the data. e. nonparametric ANOVA can be used. 4. Furthermore. For instance. We will also omit the numerical details on the results reported by the authors. and only few hints on their mathematical aspects will be given. e.71. [56]. experimental design was used in conjunction with sequential extractions to optimize the experimental conditions or to study their effect on extraction efficiency [70.1): nevertheless. A number of papers report the preprocessing of the data.71.56–59. Factors can be experimental conditions (temperature of sample treatment. The principles and applications of the chemometric techniques considered in this review will be described in the following sections. They studied the effect of organic amendments on a contaminated soil: when ANOVA revealed the presence of significant differences between samples.) [5–7]. The following examples show some possible applications of ANOVA.40–42. are referred to studies in which these techniques were used in conjunction with multivariate techniques. one-way ANOVA is performed.57. Katsaounos et al.

1. the loss of information can be expressed by introducing a matrix of residuals E: X (n × m) = R(n × m) W (m × h)T + E (n × m) (5) A rotation of PCs can be carried out. most of the papers considered in this review report the use of correlation analysis. If h PCs are retained. thus allowing an easier interpretation of the results. and vice versa.125. are: . Applications The main findings obtainable from the examination of the values of scores and loadings. Therefore. .108. e. The choice of the number of significant PC can be made with various criteria. V2 . Lucho Constantino et al. i. reflecting similar characteristics. it permits to visualize and confirm the computed correlations between variables. and the meaning of such correlation must be interpreted taking into account the knowledge on the investigated system.the influence of each variable on the PCs: variables with high loadings have a high influence on the PC and vice versa. MnO) were considered.or three-dimension plots.2.82.e. It is based on variable reduction through the calculation of the so-called principal components (also named “factors” or “latent variables”). such as adsorption and ion exchange.114]. As Table 2 shows. An example of the use of PCA for this goal is the paper by Filgueiras et al. the coordinates of the samples on the PCs.the grouping of variables into factors. observable from the combined plot of scores and loadings. or the amounts of the same element released in different fractions [47. which enable to identify at a glance the samples with high or low concentrations of some elements.g. the interpretation of the data is relatively simple and provide useful information. 5. Visualization and interpretation of experimental results The topics of visualization and interpretation of experimental results will be treated together. Therefore.wim are the loadings. when PCA is coupled to correlation analysis. usually with the Varimax method. or of the corresponding plots. PCs are not correlated with each . Fe2 O3 .110. (vi) element contents in different soil horizons [127]. yielding an increase of the weights of higher loadings and a decrease of the weights of the lower ones.1. the weights of the original variables on the linear combination. the use of biplots is discouraged by Einax et al. factors can be interpreted depending on the characteristics of the variables. assuming that the where wi1 . correlations among elements in a fraction were interpreted by Yu et al.98. (iv) extracted elements and main soil properties [35. (iii) available and total amounts [44. Principles PCA is an unsupervised pattern recognition technique. in the presence of m variables (V1 . . a technique for classifying objects into classes that are not established a priori.1.100. characteristics. The correlation between variables was used in order to study the relationship between: (i) two different elements extracted with a single reagent [22. which are linear combinations of the original variables [5–7].114 O. [6]. [34]. Vm ) the ith principal component will be PCi = wi1 V1 + wi1 V2 + . and WT (m × m) is the transpose of the loadings matrix. . [51] reported another interesting example of result interpretation: they explained the relationships among soil main properties and element contents in fractions taking into account phenomena occurring in soils. who applied the BCR sequential extraction scheme to river sediments.111. The percentage of explained variance decreases from the first PC to the second and so on. the original data matrix X (n × m). In PCA. i.113. using the commercial software packages.91.e.a classification of the objects.113.114. or conversely of a cleaner area within a contaminated site.95.95].51.g. . which remains the workhorse of many studies. The authors discussed the variable loadings on the factors.111. where n rows correspond to n samples and m columns correspond to m variables) is decomposed as a product of two matrices: X = RW T (4) where R (n × m) is the matrix of the scores.. Since the first principal components retain most of the variance..37. i.the relationships between objects and variables. (v) element contents in plants and in soil extracts [33. anthropogenic activities or natural processes. + wim Vm (3) other and altogether explain the total variance of the data.52. a correlation between two variables does not automatically imply a relationship of cause and effect between them. Finally. which represent phenomena influencing the composition of the samples. it should be recalled that the calculation of correlations is part of the mathematical treatment of many multivariate techniques. An interesting aspect of their study is that both the concentrations of extracted elements and of the sediment phase related to each fraction (e. PCA can also be considered as a technique of projection of a data set to a lower dimensional space.55. (ii) two different elements released in the same fraction in a sequential extraction procedure [48. and they could indicate the presence of a polluted “hot-spot”. CaCO3 . 5.57.116]..the visualization of multivariate data in two. or the contents of a single element extracted with two different single reagents [115] or with two different procedures [39]. 5. such as those reported in Section 5.107. or opposite.57. / Analytica Chimica Acta 688 (2011) 104–121 Obviously.104. An example of proper interpretation of correlations is the paper by Pérez and Valiente [35]. The aim of several studies on soils or sediments is the identification of the sources of elements. Examples of applications of PCA are given hereafter.117–120]. Principal component analysis PCA is the multivariate technique most extensively used for processing the results of single and sequential extractions (see Table 2). 5. On the other hand.50. or even an analytical error. The values of correlation coefficients are useful in order to make hypotheses on the sources or on the chemical and environmental behaviour of elements.the positive or negative correlations among variables.105.83. Abollino et al. which suggest their mutual influence or the presence of some common. however. many variables can be summarized by a few components [128] and a plot of the first two or three PCs enables one to visualise most of the information contained in the data. The samples with similar scores are close in the score plot: they have similar composition. in particular the differentiation between natural and anthropogenic ones.102.110. [48] as indications of a common source.117– 124]. The anomalous samples are far from the other ones.126].102. because in many cases the same chemometric technique is used for both purposes.e. This happens because it is relatively easy to apply.104. . such as chemical properties or source (anthropogenic or natural). who commented the associations between element fractions and major soil components in terms of sources or chemical behaviour: for instance the correlation of residual Pb with Fe2 O3 suggested its inclusion within resistant crystalline structures.1.

1 shows the combined plot of scores and loadings for the third fraction of Tessier’s scheme (the original sample coding was maintained). Riba et al. but not for the fourth one. La. The results of PCA showed that samples were mainly grouped according to the sampling site. studied element sources in river sediments considering element concentrations alone [48] or in conjunction with sediment phases [46. by Relic [52] for rivers sediments and by Praveena et al. Yu et al. Investigation of element sources was also performed by Abollino ´ and co-workers [38]. was similar in all sections. a few trace elements and total nitrogen: it was supposed to represent a “salinity variable”. probably because the main aim of their paper was different (see below). Bäckström et al. Cu. The top layer of the core was distinctly differentiated from all the others. whereas organic matter and Fe oxides were more accessible to elements. They just remarked the presence of common sources. Sc. Probably an examination of scores. [50] and Relic et al. chromium-plating) for Ni and Cr. a step which is not usually carried out and can be helpful for data interpretation. Cd. [58]. Pb and Zn on PC1 have opposite sign with . [49] analysed estuary and river surface sediments and a sediment profile from an area affected by the well-known and extensively studied mining spill in Aznancóllar. Pb and Zn. they can be further split into two sub groups (A17b. and traffic emissions for Pb. First of all. The presence of outstanding sampling points in the score plots was explained with the vicinity of some pollution source. The associations of the variables with the factors were extensively discussed and profitably used to hypothesize the origin and evolution of contamination. Surface (A1–A14) and vertical profile (A17b–A31) samples are in opposite areas of the plot. over time. including the removal of redundant variables. The paper by Kaiserli et al. A26. In another study [88] we examined the fractionation of elements in agricultural soil profiles from five different areas. indicating the heterogeneity of the soil material. [129]. probably because it is directly in contact with the water column. This suggests that the minerogenic component of the core. A 24 and A28. which formed the basis for the application of another multivariate technique (linear discriminant analysis) as we will describe in Part II. Spain. [51] studied agricultural soils irrigated with raw wastewater. [49]. who distinguished between mangrove lagoon sediments at high and low tide. as shown also by the paper just cited [35].g. [46. we selected the layers of the core to be treated with the BCR scheme with the aid of the results of PCA and HCA for total concentrations. who studied the fractionation of P in river sediments and its speciation within each fraction. Regarding variables. Other applications of PCA to studies on the behaviour of elements in environmental compartments are reported by El-Nemr ´ et al. Cr. Ti and Y. The latter are characterized by higher percentages of the elements of mainly geochemical origin (Sc. PCA can be used to differentiate samples according to their composition. These observations might have been made also by examining the values of the data. diffuse sources for Cd. notably to distinguish between the effect of the mining accident and of chronic pollution. A20. [56] reports another quite interesting discussion of chemometric results in terms of chemical and physical processes. industrial effluents (e.47] is somewhat simpler than those reported in the previous two papers. Pérez and Valiente [35] adopted a different approach to identify groups of samples with different element mobility and availability within a contaminated soil. and it permitted to gain insight into the interactions among the fractions and chemical forms of P. but it is anyway of interest for the interpretation of the role of phases in the binding of Co. PCA represents a useful tool also to investigate the characteristics and behaviour of elements in an environmental compartment. the separation is fairly related to collection depth. corresponding to the 190–218 cm soil layer. Our research group used PCA to characterize an industrially polluted soil [80]. Mg and Mn) and of other sediment features on the fractionation of P in lake sediments.47]. A29. so as to examine one specimen for each of the sample groups evidenced with such techniques. A remarkable aspect of their paper is the clear explanation of the criteria used in the classification. Fig. some pollution indexes (ecological risk factor and surface enrichment factor) were inserted as variables. mainly associated to such fraction. Lucho-Constantino et al. Ti. since they took into accounts scores. i. Ca. Pb and Zn: the lack of correlation between elements nominally extracted from carbonates and Mn oxides and the corresponding phases suggested that the latter were not good scavengers in the investigated rivers. The discussion of the results made by Yu et al. We will comment this figure in detail in order to show how it can be interpreted. through the values of the variable loadings. is distinctly separated from all the other samples because of the high percentages of extracted La. respectively. The top horizons of two sites markedly differed from the lower ones. An example of this application of PCA is the paper by Katsaounos et al. [52]. [22] for muddy and sandy marine sediments. anions. Sample A25. i. the influence of the most important P-binding elements (Al. with one exception (A19). The results of PCA. in addition to element concentrations and main properties.e. A21b. Finally. suggesting that they were independent of the sediment composition. PC1 was associated with electrolytic conductivity. The results of PCA can contribute to the characterization of contaminated sites. Interestingly. which were not associated to a particular matrix component. possibly because of the effect of agricultural treatments. The score values showed a differentiation between the surface and bottom samples for the first three fractions. Riba et al. alkali and alkaline earth elements. but did not identify such sources. loadings and the results of HCA (see part II). et al. suggested a separation between higher and lower sections of the core associated to a stronger fingerprint from biogenic and geological processes. these results were interpreted as a competition among various sediment phases for binding with elements. / Analytica Chimica Acta 688 (2011) 104–121 115 binding behaviour of the elements indicates the occurrence of different pollution sources and hypothesized the following sources: discharges of human origin for Pb and Cu. in addition to variable loadings. PCA provided a classification of the samples according to their contamination level. PCA was then applied to the fractionation results.O. A23. which were associated to organic matter. which had a different behaviour from the other elements and was associated to the Fe–Mn oxide content. A30 and A31). combined with those obtained for total concentrations. For example. but the results of PCA allowed us to visualize these features more easily and more rapidly. A27. indicating that the soil properties influenced the behaviour of elements towards extraction. while the more available fractions had a larger variability over the length of the core. and to the evaluation of the effectiveness of decontamination treatments. The authors investigated. Another noteworthy feature of this paper is a detailed description of the procedure used for performing PCA. Fe. using samples scores to differentiate sampling points and variables loadings to gain insight into element behaviour or sources. Positive and negative correlations among variables were also discussed.e. A remarkable feature of their paper is the interpretation of the meaning of the factors. The discussion of the variable loadings on PCs is a good example of interpretation of the results of PCA in terms of chemical and physical processes. might have given some additional information on the differences between the two lakes. Y and Al) and lower levels of the elements known (from a previous study [89]) to be main pollutants of the investigated soil. Abollino et al. the loadings of Cd. Our research group [38] studied element distribution in a marine sediment core. Ni. such as sewage discharge.

a clear separation between the extracts from the two soils was apparent. [36] used PCA and linear regression in order to assess the associations between the amounts of elements extracted Fig. such as multiple linear regression (see part II). K and Mg) having lower ones.g. [116] compared the score plots obtained for river sediments considering total and extractable concentrations respectively. the sample scores showed differences among samples associated to depth of collection. plotting variable loadings on PC2 vs. [37] studied the effect of the addition of three organic residues on element availability and main properties of a contaminated soil. 2 shows. Loadings from PC2 based on LOI-normalized data (all fractions) for roadside soil samples versus the leachable fraction (averages for all 16 samples calculated on a dry weight basis) for all elements. Variable loadings rather than scores are usually considered in these kinds of studies.51 %) 2 1 0 -1 -2 -3 A6 A14 A1 Pb Zn Cu A17b Sc Ti A24 A28 La Y A20 A21b A3 A9 A5 A7 A19 A11 pH Al Fe A27 Cr V Ni A26 A31 Mn A23 A30 A29 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 PC1 (40. Pb and Zn and an increase in pH and organic matter upon amendment addition. Al. which allowed them to well distinguish among soils treated with different amendments and to visualize the effects of the latter. respect to the other variables. which revealed the influence of urban traffic (PC1) and agricultural treatments (PC2). From [50] by permission of Springer. in our opinion. as Fig. The application of PCA to investigate the relationships between soil and biota is less common. since this subject is often studied with other chemometric techniques. after column standardization [80]. Madrid et al. / Analytica Chimica Acta 688 (2011) 104–121 5 4 3 A25 PC2 (15.11 %) Fig. the cumulative leachable fraction percentage. Pb. with anthropogenic elements (e. the elements are almost completely lined up (r2 0. [50] applied PCA to compare dry weightand LOI (loss on ignition)-normalized sequential extraction results for roadside soil samples: the variable loadings for the LOInormalized data were found to better discriminate anthropogenic and lithogenic elements than those from dry weight-normalized data. For instance they observed a decrease in the mobile fraction of Cu. In addition to the usual combined plot of PC1 vs. In a previous work we treated a profile of the same soil and one profile from another industrially-polluted soil by single extractions [89]. in a subsequent work of the same research group Tokalio˘ glu et al. . The results were not completely satisfactory. A further field of application of PCA is the evaluation or optimization of a procedure for sample treatment or data processing. [125] found significant relationships between element concentrations in grape plants and in soil single extracts using correlation analysis.76. When data from both sites were treated together. r2 = 0.76). Bäckström et al. Abollino et al. 1. In both cases. suggesting a competition between the main pollutants and the other elements for binding to Fe and Mn oxides. Bleeker and van Gestel [112] assessed element bioavailability with test organisms. PCA and HCA: it was concluded that the extractable fraction of metal was readily available to plants.116 O. Finally. A similar use of PCA was made by Maiz et al. On the other hand. PC2. as they did not observe clear associations between plant and soil contents. Alvarenga et al. Combined plot of scores and loadings obtained by PCA for pH and element percentages extracted from contaminated soil samples (coded A1-A31) into the third fraction of Tessier’s procedure. [94]: the examination of variable loadings on factors showed the presence of relationships associations between element contents in grass and in soil portions obtained with a short sequential extraction procedure developed by the authors. This finding confirms the importance of determining element mobility in addition to total concentrations. 2. Correlation coefficient.g. They thus concluded that the short procedure is suitable for studying availability in polluted soil-plant systems. and observed that the distinction between polluted and unpolluted samples was different for some samples in the two cases. The correlations between Sc and Ti and between Y and La may be due to a common geochemical origin. Cu and Zn) having high loadings and lithogenic ones (e. whereas the lack of correlation between Fe and Mn suggests the presence of two different mechanisms of dissolution of their oxides. the paper is however of interest as an example of interpretation of the meaning of PCs. the total and extractable element content and the biomass of earthworms. They made a clear interpretation of the combined plot of scores and loadings. the authors represented the data with an original approach. namely earthworms. and used PCA to evaluate the association among the soil main properties. Tokalio˘ glu and Kartal [33] used PCA to compare the element content in vegetables and in the BCR first fraction (acid-soluble elements) extracted from urban garden soils where the vegetables were grown. Another interesting aspect of the paper is the validation of results with the leave-one-out procedure [5]. Wang et al.

for instance Wang et al.132]. Common variance is expressed by the value of the communality. If the chemical behaviour of elements or the individuation of sources is of interest. atmospheric fallout. are reported when risk of pollutant release is evaluated. Examples of factors influencing element extractability in soils are agricultural practices.the simplest one is named matrix unfolding or matrix augmentation (MA). three main ways of processing data on sequential extractions were adopted in the above-cited papers: (i) the amounts (or percentages) extracted into each fraction were processed separately [33–35.49.3. such as sewage sludge.39. for instance when the contamination level is estimated.94. [39] used PCA (together with other mathematical methods) to compare the performances of two extraction procedures for an urban sewage sludge and a sludge from an olive oil factory.131]. who studied the characteristics of Alfisols. [81] adopted a column leaching approach for the investigation of tailings damp materials from gold mines and demonstrated the suitability of such procedure by FA.48. such as its solubility and the redox conditions. but also to other matrices. Three approaches were followed [5]: . (iii) the sum of the first fractions [39.2. This mode of data processing can be adopted also when the focus is sample differentiation from score values. Ultisols and Oxisols soils. atmospheric particulate matter and fly ash.116. evaluation of an experimental procedure. Furthermore. The aims of the studies using this technique are analogous those described by PCA.46–48. among leachates collected in two subsequent days or obtained with extractants at different pH.2. 5. such data were treated with multi-way data analysis techniques. such factors were interpreted by multiple linear regression (MLR) as described in Part II. the dataset can contain only element concentrations [22. then the separate treatment of the data from each fraction is advisable. [131]. who identified three biotic and three abiotic factors in forest soils. the cumulative amounts extracted in the first fractions. Handt et al. [116] found that the residual fraction was distinctly different from the other ones. i. Cukrowska et al. They processed their own experimental results together with literature data on river sediments: this approach is not so common and it could be fruitfully applied also in recent studies.52. Principles The goal of factor analysis (FA) is to find common factors explaining the experimental results [5–7. In general. [130] studied the mobility of elements precipitated as ferrite from polluted effluents.e. The inclusion of all fractions in the same PCA is necessary when differences among fractions are to be pointed out. Principles Sequential extraction results are multidimensional. which is supposed to represent the available proportion of elements. In our opinion the most proper procedure depends on the aim of the investigation.52. notably traffic density.58.2.112. i. 5.1). The score plots showed the differences among the layers of the damp. and by Illmer et al.80. (ii) the data on all fractions were treated simultaneously [36. [97]. Applications FA has been less extensively applied than PCA to single and sequential extraction results. elements and fractions and can be arranged in a three-way data array consisting of rows. In some studies.45. rainy and post-rainy seasons.94. / Analytica Chimica Acta 688 (2011) 104–121 117 by single extractions and the BCR protocol (sum of the first three steps) from urban soils. An example is the paper by Pardo et al. (ii) the separate calculation of FA for dry.80. or include main sediment or soil properties [34–37. which actually may have a great influence on sediment composition. the three-way data array X (isamp × jel × kfrac ) is unfolded into a two-way array Xaug . Arcega-Cabrera et al.45. can be applied not only to soils and sediments.e. Abollino et al.88.1. i.56. based on the relationships between the bioavailable element portion and the main sediment properties.129.52. a decontamination technique is tested or a procedure for sample treatment or analysis is evaluated. Various criteria can be used to choose the number of significant factors and disregards the last ones.131]. The paper by Yun et al.51. Two remarkable aspects of their paper were (i) the discussion of variable contributions to factors. mainly regarding PCA. this is not indispensable in some applications. After factor rotation. Pérez Cid et al.O. for a better characterization of the matrix under study.133]. In FA the data matrix X is decomposed as: X = FLT + E where F is the score matrix. 5. so the authors concluded that dilute HCl was unlikely to be a suitable alternative to BCR procedure to estimate potential mobility and extractability of elements. Pardo et al. are defined by samples.1.2. [53]. showing a distinction between magnetic (crystalline) and non-magnetic (amorphous) samples from PCA scores and a differentiation of Cr from all the other elements from loading values. Whereas it is always advisable to take main properties into account. The interpretation of variable loadings on factors was in agreement with the results of ANOVA (see Section 4. A few examples are given here. The meanings of the factors were related to properties of the damp material. and is not a real multiway method. Obviously PCA.56. Even if most applications of multivariate techniques to extraction assays were published since 2000.116].47. some papers on this subject appeared before that date. [54] used FA to gain insight into the geochemical processes involving Pb in river sediments. [111] on contaminated roadside sediments is an example of application of FA to the identification of sources and of causes of spatial and temporal distribution of elements.3. [45] utilized the trends of the variable loadings to support their hypotheses on different sources of elements (industry and traffic) in dust samples. in order to study the temporal variations of Pb geochemistry.130. Other examples of application of FA are the papers by Dalurzo et al.89. It is interesting to note that the authors included some non-chemical variables in the data-set.38. 88. suggesting the different mobility of elements bound to it.33. The equivalence between the two methods was judged to be incomplete.38. percentage of metal industries and monthly precipitation.46. the nature of the variables most heavily contributing to each factor enables to identify its meaning. LT (6) is the transpose of the loading matrix L and E expresses specific feature variance and residuals. columns and layers. Finally. characterization of chemical behaviour of available elements.125. who discussed the values of loadings and scores for the residual fraction of Tessier’s scheme in river sediments. assuming that such fraction is smaller in polluted than in unpolluted rivers. as well as the other chemometric techniques considered in this review. identification of sources.58. the part of the variance of one feature which is described by the common factor solution in the factor analysis.e. The total variance of the data is divided into three parts: common feature variance. in order to extract more information than that obtainable with bidimensional methods [40–42.50. Factors represent the com- mon variance of features. which are supposed to express noise. Factor analysis 5.49–51.116].130]. N-way methods 5. specific feature variance and residuals or errors.

who compared the performance of MA-PCA.e. The authors chose the model with complexity [2431]. A second issue is the interpretation of the outcomes of the three techniques. C(k × R): P Q R xijk = p=1 q=1 r =1 aip bjq ckr gpqr + eijk (9) where P. elements and 2 represents fractions) and gpqr are the elements of matrix G. among elements (B mode) and among fractions (C mode). and surprisingly did not reveal a difference between measurements in inert and non-inert atmosphere. but the use of this technique will probably increase in the future. The choice of the value of r is critical and can be made with various criteria [40]. a condition which was not fulfilled with their dataset. so that r xijk = f =1 aif bjf ckf + eijk (8) where r is the number of factors retained and eijk is the residual error term. [133]. Cu.e. . In addition. Tucker3 was extended to work with a four-way data array (samples from different depths × elements × fractions × conditions). In a previous work Pardo et al. the coordinates of the sampling points. They applied the BCR protocol in two conditions. which led to a characterization of the investigated soil in terms of distinction among sampling points. Tucker3 method is based on the decomposition of the matrix X into a three-way core matrix G and three two-way loading matrices A(i × P). Alvarez et al. Abollino et al. The fourth mode (D) included only one latent factor. Tucker3 method gave further information through the core array. because it requires a superdiagonal core matrix G. Ni and Zn). in which MA. A2 showed the classification of the samples according to the depth of collection. so that each element xij can be expressed as: r xij = f =1 uif vif + eij aug (7) where uif and vif are the elements of the scores and loading matrices Uaug and V. Again. Three-way PCA was more suitable than MA-PCA to interpret the three-dimensional data set (A = samples × B = variables × C = depth).’s paper [41]. PARAFAC and Tucker3 in assessing the status of soil profiles irrigated with wastewater. [40] used PARAFAC to compare the efficiency of three fractionation schemes for the characterization of estuary sediments. and direct accumulation in deep ones. Finally. The most important one explained the interaction among the first factors in all modes: its meaning is that in the top layers the analytes are present in immobile forms. Unfortunately. the three modes provided information on sampling points. Applications Presently there are not many application of N-way methods to extraction results. elements and fractions. i. 5. [136] applied MA and PARAFAC to sequential extraction results for marine sediments. and reported a contour map of the loading values in the physical space. Then the authors considered the core array: as an example. B(j × Q). owing to its potentialities. Tucker3 and PARAFAC were applied to the results of BCR scheme for soil samples The first one is a comparison among the three chemometric techniques. [42] assessed the performance of Tucker3 method using data on two soil depth profiles collected from a contaminated site. The third mode (C) was related to the fractions. hence the plot of A1 vs.2.135]. The authors concluded that the four-way Tucker model chosen enabled them to obtain some information on bioavailability and behaviour of elements. the information on samples and fractions is confounded with this procedure. Q and R are the numbers of factors assigned to each mode (in this case. four factors in the second one and so on. An advantage of Tucker3 method over PARAFAC is that different numbers of factors can be used for the different modes. we will report the meaning of two elements of the core. the paper combines a rigorous data treatment with a proper interpretation of the results in terms of environmental and geochemical processes and is a good example of the potentialities of the application of chemometrics to extractions. showing the most potentially hazardous areas. which showed that the information obtainable increased in the order MA < PARAFAC < Tucker3. dedicated much attention to the procedure followed for the calculations or compared the performances of more data treatment techniques. modes are information about samples. B = variables. with two factors in the first mode. the second most important core element was related to the trend of residual Cd with depth. In our opinion. / Analytica Chimica Acta 688 (2011) 104–121 ((isamp × kfrac ) × jel ) which is then treated with the conventional procedures of PCA. Fig. namely inert and non-inert atmosphere. Similar findings were made with both PARAFAC and Tucker3 on (i) differences among sampling sites (highlighting the effect of the distance from the wastewater outlet).Another approach is parallel factor analysis (PARAFAC). for instance deriving from a single extraction on different layers in a soil profile [133]. showing the relationships among sampling points (A mode). they found that samples could be classified according to their Tucker3 Amode loadings. they could not use Tucker3 method because the core matrix G was not sufficiently superdiagonal. . Three interesting aspects emerge from Pardo et al. i.3. 3 reports the loading plots for the three modes. C2 and C1 vs. The first mode (A) was related to depth. the authors plotted A-mode loadings of factors 1 and 2 vs. Cr. An example of application of N-way analysis to single extraction results is the paper by Singh et al. B (j × r) and C (k × r). which is based on the decomposition of a three-way data array X (i × j × k) into three loading matrices A (i × r). whereas PARAFAC covered all the three aspects. gpqr the interactions among the three modes. MA provided information on elements and fractions but not on the global behaviour of sampling points. The good performance of Tucker3 method was due to the possibility of choosing different numbers of factors for each mode (A = samples. N-way methods can be used also with other kinds of multidimensional data. Cd and Ni) and consequently different environmental hazard. For instance. C = fractions). B4) differentiated them in two groups. (iii) differences among layers with different contamination levels. description of element behaviour and of discrimination between predominant and minor fractions. Stanimirova et al.118 O. Like Pardo et al. A method of re-folding the matrix in order to separate the information has been proposed [134.Finally. it was observed that sampling sites receiving highly . B2 and B3 vs. it is not surprising that some papers. The second mode (B) was related to the elements and two projections of the four factors (B1 vs. C3 showed a clear differentiation between the first three fractions and the fourth one. due to different accumulation pathways: dust precipitation and the flooding of a small stream in the top layers. They could not use Tucker3. The following examples will help clarify the mechanism of multiway analysis. in addition to studying the characteristics of the samples of interest. the authors searched for a single parameter that could model the potential environmental hazard of individual soils. the plots of C1 vs. (ii) elements and main soil properties (identifying wastewater-related elements. obtaining two contour plots which show areas having different levels of available toxic elements (As. [136]. Due to its being relatively “new”. r is the number of components retained and eij is the error term.

189 (2008) 291–333.W. in order to improve comparability among results. In general. Both Tucker3 and PARAFAC have advantages and disadvantages. A. Chem. 21 (2002) 451–467.O. Weinheim. Geoderma 117 (2003) 347– 366.g. Acta Hydrochim. 1997. Loading plots obtained with multi-way PCA applied to the results of the BCR scheme for marine sediments. Wiley-VCH. (a) A-mode. so a harmonization of the chemometric data treatment procedures would be desirable.G. Cambridge.W. As the need for harmonization of experimental procedures led to the BCR sequential extraction procedure. S. Bacon. and we presume that its applications will increase in the near future.M. R. different ways to calculate PCA (e. [2] C. mostly at depths between 5 and 20 cm. López-Sánchez. (b) B-mode. Fig. Handbook of Chemometrics and Qualimetrics. 6. contaminated wastewater were adversely affected in terms of main soil properties. It is important to have a good knowledge of the characteristics of the matrix and of the properties of the elements under study. Elsevier. F. Terribile.g. Quevauviller (Ed. its main applications are the investigation of the sources of elements and of their biogeochemical behaviour. B. Parts A and B. L. PRIN project) and the Italian National Research Program for Antarctica (PNRA) for financial support.R. Adamo. in order to correctly interpret the outcomes of data processing and obtain as much information as possible from them. the application of more than one technique to experimental results can be useful to confirm or exclude an hypothesis or to examine data from different points of view. Trends Anal. VCH. / Analytica Chimica Acta 688 (2011) 104–121 119 core array. Water. Zwanziger. M. Hydrobiol 34 (2006) 425–436. The above examples show the potentialities of N-way analysis for the treatment of multidimensional data. Rome. Vandenginste. Bermond. using commercial software packages. J. S. [6] J. Geiß. Soil Pollut. Swennen. Acknowledgement We thank the Italian Ministry of Instruction. However. 9 (2007) 319–328.M. Denaix. Kowalik.C. (c) C-mode. Einax. Einax. H. Acta 559 (2006) 105–112. [5] D. PCA is the multivariate technique most extensively used with these kind of data. Massart. Zampella.J. Smeyers-Verbeke. Chim. De Jono. J. Manouchehri. the identification of areas with different contamination levels and the assessment of experimental or computational procedures. Besancon. PARAFAC gives a less complete picture of the investigated samples. with or without Varimax rotation). Tellier. Astruc. P. but its results are easier to interpret than those of Tucker3. The above cited paper reported different data preprocessing or standardizing procedure (or did not mention these steps). Buydens. analytical chemists and expert in the investigated matrices would be very useful in order to apply the techniques correctly as well as to obtain as much information as possible from the experimental results. Davidson. since the core array must fulfil some requirements. S. normal distribution) are known and respected. Air.M. Weinheim.). Otto. owing to their capability to deal with multidimensional data. S. it is important that the meaning and the steps of a technique are understood before its application. Royal Society of Chemistry. [9] P. From [136] by permission of Elsevier. 1997.R. J. The variable reduction techniques described above can be adopted also by researchers who are not expert in chemometrics. Sahuquillo. Monit. Rao. thanks to the possibility of choosing the number of factors for each mode and to the presence of the . [10] V. Anal.W. Environ. University and Research (MIUR. Methodologies for Soil and Sediment Fractionation Studies. 3. We presume that the application of N-way methods will increase in the near future. A cooperation between chemometricians.L. Amsterdam. Leqy. References [1] J. Gleyzes. 2002. [4] C. The former is more informative than the latter. [3] P.F. Cappuyns.M. [8] C. J. C. M. and its assumptions (e. L. choosing the number of significant factors and interpreting the results. [11] N. [7] M. Abollino et al. Analyst 133 (2008) 25–46. A. Chemometrics in Environmental Analysis. but it is not applicable to all datasets. Chemometrics: Statistics and Computer Application in Analytical Chemistry. Concluding remarks Chemometrics is a powerful tool for the treatment of single and sequential extraction results. TrAC. 1999. examples of studies adopting two or more chemometric methods will be reported in Part II of the review.

Wennrich. 86 (2007) 121–129. Chemosphere 69 (2007) 69–78. A. L. P. Pérez. Aten. Panda. Environ. E. Pollut. Sauerbrey. E. H. D. I.J. Chomchoei. Jardine. 363 (2006) 28–37.F. N. Y. Analyst 122 (1997) 1211–1221. D. A. Chemosphere 74 (2009) 1292–1300.C. Petrella. H. J. Biswas. [80] O. L.H. E. [90] P. Can.M. Rauret. Chim. C. Acta 536 (2005) 183–190. Wragg. – c.120 [12] [13] [14] [15] [16] [17] [18] [19] [20] O.E. Sci. Anal. Soils Sediments 8 (2008) 415–423. Vázquez. Ho.E. Förstner. [83] C. Govender. S. R. Rodrigues. Wenzel. Santamaría-Fernández.J. Sci. [92] W.M. [94] I. ˜ [71] R. Sci. Agrochimica 6 (2007) 319–328. Miró. Environ. Beer.C. P. 40 (2009) 376–390. Barberis.R. M. E. Intell. A. Matúˇ s. Chem. 330 (2004) 115–129. E. Giacomino. 148 (2003) 3–14. R.B. B. Ahumada. J. I. J. Illmer. Geochem.D. Part A 36 (2001) 971–986.W. E.I. 51 (1979) 844–851. Dinelli. Friel.G. Sarzanini. (Rome) 95 (2005) 525–538. Kalahne. H. [77] E. Debán. A. Mar. Díaz-Barrientos. 83 (2003) 935–952. T. 17 (1998) 632–642. Malandrino. Álvarez-Suárez. Anal. 39 (2007) 2429–2435.L. Madejón. Analyst 126 (2001) 1304–1311. Xu. G. L. Gallardo. Weisz. Daesslé. TrAC. Soil Sci.M. C. B. 90 (2004) 135–160. Hall. T.W. Explor. Water Sci. Toxicol. Ruby.J. Covelo. A. Total Environ. Mentasti. Yang. Environ. Health A36 (2001) 1–16. S. T. Health 27 (2005) 121–130. J. Toxicol. Beltrán-Hernández. Bioanal. Environ.N. Plant Nutr. Chem.M. Chinnathangavel. Arambarri. R. Garrido.V. Naidu. L. Casteel. [98] C. R. Covelo. Technol.M. Environ. Ann. P. C. Velayutham. R. 129 (2007) 151–168.A. V. 24 (2005) 759–771. Howard. Basta. M. Ecol. Int. Hossack. A.K.A. Astel. Jordan. Mendoza. Ajmone-Marsan. US Government Printing Office. J. E. Grˇ cman. E. 31 (2005) 313–323. Vega. Mater. T. Galláraga. K. SW 846. Atmósfera 21 (2008) 335–345. Bioanal.T. Chim. Romero. E.L. Mentasti. Tokalio˘ glu. Test Methods for Evaluation of Solid Waste. Bendicho. Handt. Armienta. Water Res. Chim.J. Chang. Jiang. A. Goovaerts. Prohaska. Sobottka. M. [103] P. S. Álvarez. Cave. A. Environ. Buszewskl. [75] Y. Sci. A. Gallardo. Novillo. Massart.T. [85] M. 22 (2003) 457–465. 35 (2004) 1615–1634. Chemosphere 56 (2004) 39–50. J. Commun. 1995. Life Sci. M. Hansen. Monit. [107] C.I. [99] I. Rees.C. Sârbu. Fedotov. J.M. Lucho-Constantino. Eng. D. L. Luo. Geochem. Explor.E. 42 (2000) 193–199. B. Sci. Saha. A. Total Environ. D. M. [91] J. Meuboom. Fay. Fernández Gómez. Madrid. Abollino. P. Fendorf. El Sikaily. P. Grøn. Acta 565 (2006) 63–72. M. 50 (2005) 276–290. Shu. Grennan. A. [109] A. [88] O. Welch. Acta 504 (2004) 217–226. Norvell. J. Lin. [89] O. Garcia. S. Barnett. Chim.A. 77 (2002) 191–207. Fernández Alborés. Environ. Streˇ sko. D. Choi. Zhuang.D. Bekele. A. González. Shi. Geoderma 132 (2006) 450–463.M. R. S. I. Anal. K.C. Khaled. A. Qual. Bioanal.T. Prause. Chem. Health Part A 42 (2007) 1303–1315. Biochem. Shaw.Z. [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] . Lee. Environ. A. Chemosphere 49 (2002) 545–557. Cabrera. [95] G.W. Barnett. Arch.R. Environ. [68] B. Yu. Environ.C. 5 (2003) 929–934. Sci. Juhasz. A.S. Res. C. Environ. Bengtson. Adhikari.J. E.T. Chen. Madrid. C. J. Chem. Allard. M. Floras. Environ. L. Acta 478 (2003) 111–118. Acta 439 (2001) 139–151. M. Chim. W. S.K. J.M. Anal. Zehl. Karlsson. J. Environ. F. I.H. K. [81] E. J. Pollut.C. Chim. G. Moirou. Monit. Environ. S. Cukrowska. M. [87] R. 163 (2009) 632–644. L. Ho. [105] F. Chemom. [97] P. J. Xenidis. Zhang. J. Tack. Gallardo. ´ T. Mentasti. Commun. J. 21 (2000) 989–1000. Environ. Technol. M. Gallardo. Stark. M. G. Hazard. Santamaría-Fernández. da Costa Duarte. T. A. M. D. [110] X. Moreda-Pineiro. M. J. Miró.M. 23 (1994) 987–999.H. Chojnacki. Hursthouse. For. Rosende. Basiliko. ˜ -Meléndez. J. A. Tsai. Contam. Escudero.A.K.M. D. [72] R. Acta 420 (2000) 205–216. [82] B. 35 (2001) 2417–2428. 39 (1990) 197–208. C. M. E. Katsaounos.J. Aceto. E. A. U. [76] A. Snape. [64] J. Voutsa. Filgueiras. Spivakov. 36 (2002) 4562–4569. Strawn. Plant Soil 311 (2008) 109–119. Palumbo-Roe. Soil Sci. Samara. B. O.H.J. Keller. E.I. Technol. Water. Commun. C. C.A. 148 (2007) 824–832. J. Davidson.H. [108] E. Soil Sci. Geoderma 137 (2007) 352–359. Chem.C. M. Chojnacka. Environ. Y. Sci.J. Sahu. A. Lulli. Sansom.M. D. M. Qual. S.T.B. Environ. R. S. J. Chem.S. El Nemr. A. M.E. 147 (2007) 713–722. Miró. Environ. Environ. J. W.M. 10 (2008) 11–25. Environ. B.P. M. [84] R. Environ. Pérez de Mora. Koelmans. Yang. J. Eng. V. Kubová. [70] M. Sutherland. E. Santamaría-Fernández. Urquhart. Acta 658 (2010) 41–48. S. G. 291 (2002) 123–134. P.C. Environ. Grayston. Marcano. Cabrera. Duarte. B. Chem.P. J. Riddle. [104] P.Y. Delvalls. Z. A. Eberle. US EPA.V. Medved’. Sarzanini. Riba. 394 (2009) 337–349. Total Environ.J. 44 (2003) 281–287. O. Davidson. Leinweber. Anal.S. Campbell. Ollesch. Talavera. E. A. Heilmann. Quevauviller. [112] E. Hansen. G. Bailey.M. [102] A. 257 (2009) 739–746. Environ. M. Soil Sci. J. M. Assess. Giokas. Fernández-Palacios.A. F. 7 (2007) 170–176. 7 (2005) 29–36. Evans. Bäckström. F. R. A. Int. R. C. G. Pereboom. 119 (2002) 177–193. I. L. P. J. Anal. Bhatta.C. Reinoso.M. Wu. Dalurzo. Ðordevi ´ A. Shiowatana. third ed. A. Gupta. Environ. Oomen.G. Rofe. E. 4 (2002) 330–336. U.Y. Sci. Valiente. E. A. 152 (2009) 133–147. Tessier. E. Tsai. Bisson. Lopez-Valdivia. Giacomino. [73] M. J. Manage. M. Williams. Environ. A. A. Alvarenga. Fernández Band. Davidson. Stark. Dótor. M. [114] A. Pollut. Arcega-Cabrera. Sci. V. Plant Anal. Cerdà. F. [101] J. 40 CFR Parts 403 and 503. [79] O. F. Geoinf. Bull. Malandrino. A. Snape. / Analytica Chimica Acta 688 (2011) 104–121 P. DC. Assess. ´ D.A. de Varennes. López-Sánchez. A. Abollino et al.A. M. S. Klinck. E. 8 (2003) 37–50. Vieira. M. Chem. Wang. Int. S.R. Cave. J.T. J. Kartal. Védy. K. M. Toxicol. Bleeker. Nathanail.M. Anal. C. Environ.H. [66] B. [86] P. S.G. A. 36 (2005) 1499–1511. Pollut. Popovic. M. Yun. Madrid. C. Burgos. Monit.H. Furman.J. Sci. Qual. Malandrino. G. Geoderma 151 (2009) 303–310. L. M. I. C. Laboratory Manual Physical/Chemical Methods. Higgins. 337 (2005) 175–182.G. J. 137 (2006) 315–338. Meiner. Mentasti.M.G. J. D. [93] M. Health. Boszke. I. 91 (1996) 89–96. E. Abollino. Ho. Cazurro. Sandrini.J. C. A. Chem. Chem. M. H. A. Environ. Pérez de Mora. Adv. Vander Heyden.M. J. S. Geochem. Oomen. Heinz.A. Wragg. Aceto. Durán. K. J. Trends Anal. Geoderma 98 (2000) 95– 113. Part A 42 (2007) 1251– 1261. P. A. Pollut. S. [111] S. Lista. Anal. Sci. M. 110 (2000) 3–9. M. M. A. Carretero. Gore. Zehl. Castrignanò. Total Environ. Gonc ¸ alves. R. Sellstone. 30 (1996) 422–430. F. K. Malandrino. H. Assess.J. Obrador. Boonjob.J. Stouraiti. F. [106] J. J. Total Environ.J. M. Cave. 41 (2007) 406–418. S. Rath. Fernández Lopez. G. [69] J. Giacomino. Leonardos. Kowalkowski. Analyst 131 (2006) 509–515. Amin. Frenzel. Plant Anal. E. W. Bragato. G. J.F. Aceto. Vallini. J. Chen. J. Environ. Soil Biol. Kersten. Air. K. Int. P. M. C. Assess.J. Dimirkou. van Gestel. Geoderma 134 (2006) 187–199. Ascar. Washington. M.C. Wragg.G. O. Chim. Monit. Cave. Díaz Barrientos. H.J. J. R. 158 (2010) 416–423. I. G. R. Relic. Environ.Y.E.R. R. Fernández. Water Sci. S. Anal. R. Earth Obs. J. S. Environ. A. 109 (2000) 497–504.O. Madejón. Stark. E. C. P. Carton. Anal. Environ.R. O’Donnell. Anal. F. 155 (2009) 205–213.A. B. 379 (2004) 108–114.C. P. Giacomino. Davis. J. Zupan. Yu. Munoz [96] Z.J. Environ. J. Kuchel. K. M. Soil Pollut. J. S. Weber.H. 35 (2006) 450–458. IA. P. Appl. Intern. Monit.W.J. Savonina. M. Acta 620 (2008) 34–43.E. C. Riddle. Anal. L. Hill.M.F. Yu. 92 (2010) 453–475. M.E. Paz González. A. Environ. Gómez-Parra.M. Monit. Technol.G. Pure Appl. Fernandes. 27 (1998) 335–342. 18 (1986) 121–130. Abollino. R. TrAC. 18 (2002) 353–362. Parry. Tsai. Falqué López. M. Lavilla. X. Plant Soil 280 (2006) 7–21. Taylor. D. M. Chemosphere 46 (2002) 1147–1155. van Griethuysen.A. Monit. [113] A. Häni. Barberis. J. Cave. Acta 606 (2008) 26–36. Klinck. Plant Soil 273 (2005) 269–277. A. [100] A. M. Health. Miró. Einax. 81 (2001) 469–479. Scouller. [65] J. C. Blagojevic. M.K. R. Hudnall. Anal.W. Hlavay. Syst. Mentasti. Bruil. 24 (2009) 162–171.A. Biasioli.. P. F.K.C. Hill. Technol. Pérez Cid. S. Zhang. D. Sci. Castillo-Blum. A. Geoderma 146 (2008) 378–390. Chem 385 (2006) 771–779. Rompelberg. Abollino. Trends Anal. Viljoen.C. A. Fernández-Palacios. Pueyo. Plant Anal. Einax. Cerdà. Smith. 23 (2006) 589–595.E. L. vol. Dobbe. Anal. Cave. Palumbo-Roe. C. Kaiserli. Environ. Stanimirova. M. J. Hill. Anal. Soil Use Manage. J. Alvarez. Górecki. ´ Environ. 168 (2005) 307–315. R. Bujdoˇ s. Karayannis. [67] A. Y. Water Res. Chuna-Queda. Chemosphere 57 (2004) 491–504. M. I. Maiz. Lafuente. Vazquez. H.J.L. A.L. F. R. Taboada Castro. Pardo. Górecka. Hansen. Pollut. Rodríguez. K. Jardine. Malandrino. K. J. Georgiadis. Monit.S. M. N. Rodríguez. Environ. Ho. Rico. Chang. Milodowski. C. Gilmore. Anal.S. L. Lab. Hoashi. Wang. L. Huang. T. J. R. Jayachandran. K. Chim. F. Cabrera. Forja. B. O. A. Golia. Technol. 56 (1996) 59–78. J. González. Chim. Assess. C. P.L. C. Gómez. C. M. M. K. Schinner.C. Benzo. 31 (2005) 661–669. Rodríguez. Shi. Christie.M. Durán. J.B. S. Chomchoei. Prescott. Abollino. Acta 557 (2006) 344–352. Pollut. 178 (1996) 45–53.T. W.O. A. Obertegger. Mentasti. A. J.K. U. Florido. Chim. S. McGrath.L. S. Feng. Ajmone-Marsan. M. Moreno. Abollino. 76 (2004) 415–442. Quintana. F. O. S.A. Stingeder. Geochem.K.W. Rosende. Burgos. [78] J. C. A.M. Sips. C. Sci. Soil Sci. S.Y. Pollut. M.J. Vaive. [74] R. 4 (2004) 71–86. Stark.R.A. Millán. G. Korre. Schoof. Environ. J. Richter. Malandrino.E. Chim. Tositti. Cai. E. R. Edelgaard. 39 (2005) 7102–7110. Mossop. L. A. Water Air Soil Pollut.R. Anal. M. K.

A. Pardo. 17 (2005) 540–544. Meers. J.O. Barrado. Barceló. M. J. Y. J. C.M. E. H. 2 (2008) 139–148. Geebelen. B. 132 (2006) 80–89. Zhang. R. Aust. Yanai. Vega. Schnug. S ¸ .Y. Liu. Smith. M. Lab. Soil Sci. Soil Sediment Contam. Anal. Res. Pardo. Singh. Anal. Plant Anal. Environ. / Analytica Chimica Acta 688 (2011) 104–121 [115] E. [123] Z. Anal. Geoderma 97 (2000) 39–60. Ruttens. [126] G. C. M. D. 55 (2009) 110–120.J. Ren. G. L. Fleckenstein. K. Aris. [128] S ¸ . Malik.A. Barceló. L. M. Vandegehuchte. [135] M. Meers. Geoderma 141 (2007) 247–259. G.R.H. Risaliti. [124] Z.S. Int. Intell. N.L. Inorg. Zhang. J. Song. C. Soil Sci. Jin. Plant Anal.M. [129] S. B. Yang. Int. B. Chem. 13 (1999) 260–287. Kartal. Sánchez. Stochastic Environ. Abdullah. 34 (2000) 3307–3314. Environ. Li. Environ. Abollino et al. A.P. 17 (2008) 467–485. Tauler. D. J. Naidu. M. Radojevic.H. Chim. A. Song. J. K. Mater. Aydin. Qual. 83 (2006) 1–12. Debán. F. X. C. Tack. Hazard. Mladkova. Recatalá.M. Samson. Martinelli.W. [120] C. Kartal. A. Ruttens. Tack. 50 (2004) 565–573. Diels. Moritsuka. Zhang. Risk Assess. Korre. Anal.M. Unamuno. Syst. [117] A.-B. Tokalio˘ glu. Pardo. [118] N. Bruland. Zhang. Peris. E. E. Castrillejo. Vega. Thurman. C. X. Soil Sci. Z. [132] A. Lett. Geoderma 136 (2006) 107–119. Chemom. J. [121] A. 40 (2009) 2389–2412.W. Ren. Soil Res. [131] R. L. [136] R. Singh. Verloo. 33 (2004) 785–794. V.L. [122] G. Vangronsveld. Velasco. Drabek. V. Plant Soil Environ. Terrado. D. 33 (2002) 3347–3357. L. M. J. Commun. C. M. Res. S. [130] R. Du Laing. Unamuno. R. F. Castrignanò.Z. 121 ˝ [125] S ¸ . V. S.R. Environ. Wang. Tauler. Vangronsveld. S. M. J. Ahmed. Sci. Technol. Commun. R. Y. 99 (2005) 1796–1806. Haneklaus. Hu. Vega. J. 20 (2002) 187–195. W. Umeda. 84 (2004) 691–705. L.A. Wang. E. Sinha. C. Richardson. Guerra. Giugliarini. 26 (1993) 1719–1739. Praveena. Quim. Plant Nutr. .M. T. Acta 523 (2004) 125–132. Boruvka. A. [134] R. Castrillejo. M. Tokalio˘ glu.-G. Liu. Kosaki. S ¸ . R.K. J. Environ. [119] E. O. Micó. L. Talanta 70 (2006) 691–704. Zhang. C. J. Luo. Wang. M. Helena. Z.H.S. Liu. Gassner. F. K. L.G. Yuan. Gunes ¸ . Biochem. S. [133] K. 38 (2000) 823–836.G. Guerra. G.Y. L.A. Jiang. Cazurro. Krishnamurti.G. A. Sci. D.M. Prieto. [116] Y.M. Barrado. Vanbroekhoven. [127] L. Du Laing.

Sign up to vote on this title
UsefulNot useful