Preconcentration and flow-injection multivariate determination of priority pollutant chlorophenols

Preconcentration and flow-injection multivariate determination of priority pollutant chlorophenols

ANALYTICA CHIMICA ACTA Analytica Preconcentration and flow-injection multivariate determination of priority pollutant chlorophenols F. Navarro-Vill...

626KB Sizes 0 Downloads 56 Views

ANALYTICA CHIMICA ACTA Analytica

Preconcentration

and flow-injection multivariate determination of priority pollutant chlorophenols

F. Navarro-Villoslada, Departamento

Chimica Acta 308 (1995) 238-245

L.V. P&-ez-Arribas, M.E. Lecin-Gonzglez

de Quimica Analftica, Fact&ad de Ciencias Quimicas, Unicersidad Complutense,

*, L.M.

Polo-Diez

28040 Madrid, Spain

Received 4 May 1994; revised manuscript received 4 July 1994

Abstract A combined flow-injection analysis and multivariate calibration method is developed for the simultaneous determination of priority pollutant chlorophenols (2-chlorophenol, 4-chloro-3-methylphenol, 2,4-dichlorophenol, 2,4,6-trichlorophenol and pentachlorophenol) in water. Chlorophenols were preconcentrated in an XAD-4 adsorbent resin. The ion pairs formed with tetrabutylammonium at pH 9.1 were extracted in chloroform and detected in a diode-array spectrophotometer in the wavelength range 200-430 nm. Three multivariate calibration methods, classical least squares (CLS), Kalman filter and partial least squares were used for comparative purposes. Comparison of the standard error of prediction shows significant differences for the determination of 2,4,6-trichlorophenol only, the best results being obtained by the CLS method. Keywords:

Flow injection; Multivariate

calibration;

Chlorophenols;

Preconcentration;

1. Introduction Chlorophenols are important pollutants in the environment because of their high toxicity. They are extensively used as pesticides, herbicides and fungicides and their use could be expected to leave residues in water, soil and food. Most of them are on the U.S. Environmental Protection Agency priority pollutant list and, therefore, rapid, accurate and sensitive analytical methods are required for identification and determination of these compounds in different sample matrices. The total phenol content has been determined by UV-visible spectrophotometry using different chromogenic reactions in the batch mode [I] and by flow-injection analysis (FIA) [2]; in these methods

* Corresponding

author.

0003-2670/95/$09.50 0 1995 Elsevier Science B.V. All rights reserved SSDI 0003.2670(94)00412-9

Waters

standard mixtures of phenols are proposed as standards for the analysis of different samples. Many methods for their individual determination have been proposed, mostly based on chromatographic techniques. Liquid chromatography (LC) yields limits of detection in the ng ml-’ range but derivatization [3,4], solid-liquid extraction 151 or selective extraction [6] steps must be used, Gas chromatography (GC) yields similar limits of detection to liquid chromatography (LC) and recoveries averaged better than 80% [7] and between 63-96% [8] using different preconcentration adsorbents; solvent extraction and derivatization (often with toxic derivatizing reagents) steps are also involved [g-lo]. Other techniques include supercritical fluid chromatography (SFC) [ 11,121 and capillary isotachophoresis 1131. Although chromatographic and electromigration techniques allow highly efficient separation of a large number of compounds in a single run, they are rela-

F. Nacarro-Villoslada et al. /Analytica Chimica Acta 308 (1995) 238-245

tively time consuming, with the typical sample frequency not exceeding several samples per hour. An interesting alternative for the determination of phenols is based on multicomponent analysis methods; the least-squares linear multivariate regression method has been used in the determination of individual phenols in the batch mode [14,15] and in a continuous extraction kinetic process [ 161. The limitation of FIA to the simultaneously determination of several analytes [17] could be overcome by using multivariate analysis. Of the approaches to FIA multi-determinations those that utilize multidetection systems, e.g., photodiode array spectrophotometers, offer the greatest potential, particularly for process monitoring. Multicomponent analysis with a FIA multidetection system involves two essential aspects, the recording of the spectrum under conditions of maximum sensitivity and reproducibility, and the mathematical treatment of the information. Related to the second aspect, the development of multivariate calibration to resolve the spectra of the mixture produced by a flow-injection system has become of great interest in the last few years and has been applied mainly to binary FIA multideterminations [l&21] but also for up to four components [22]. Despite the relatively small number of papers dealing with the simultaneous determination of several components by FIA, this very effective technique has increasing importance for routine practice and research. Several authors [23,24] pointed out the need to compare different multivariate calibration methods because it is impossible to know a priori which method gave the best results. They concluded that because chemical data sets tend to be different, there is no reason to expect that the best method on one data set is going to be the best for all other data sets and the best method is the one which gave the best predictions. A comparison of multivariate calibration methods for the simultaneous determination using a double-injection FIA approach has been reported [21]. Within the multivariate calibration methods, direct methods [24,25] have been successfully applied in FIA but the use of recursive methods [26,27] has not so far been published. This paper describes the development of a method for the preconcentration and multicomponent determination of five priority pollutant chlorophenols in water by a combined FIA system with photodiode

239

array detection, derivative spectrophotometry and data treatment using three different multivariate calibration methods for comparative purposes. The methods applied were classical least squares (CLS), partial least squares (PLS), a biased method (both are direct methods) and a Kalman filter, a recursive method.

2. Experimental 2.1. Apparatus The FIA set-up consisted of a Tecator FIAstar 5020 Analyzer that includes two peristaltic pumps and a variable volume Tecator L-100-1 injection valve, a membrane phase separator, a Hellma quartz flow cell of inner volume 18 ~1 and optical path length 1 cm and an HP 8452A diode-array spectrophotometer interfaced to an HP Vectra computer and an HP Think Jet printer. All coils and sample loops of the manifold were constructed from PTFE tubing and the solutions were propelled in PVC pump tubing. A WIZ peristaltic pump was used in the preconcentration step. 2.2. Reagents All chemicals were of analytical reagent grade and purified water was obtained from a Milli-Q purification system. Chlorophenol reference standards were obtained from Carlo Erba and Aldrich. Stock solutions of 2chlorophenol, 4-chloro-3-methylphenol, 2,4-dichlorophenol, 2,4,6-trichlorophenol and pentachloropheno1 (100 mg ll’) were obtained by initial dissolution of accurately weighed substances in 1 ml of 0.1 M sodium hydroxide. Working solutions were prepared by suitable dilution of the stock solution with water. A 1500 mg 1-l stock solution of tetrabutylammonium nitrate (TBA) from Fluka was also used. The pH was adjusted with a 0.01 M borate buffer at pH 9.1 in 0.05 M NaCl to fix the ionic strength. Spectrophotometric grade chloroform from Carlo Erba was used in the extraction. The macroreticular resin XAD-4 from Rohm and Haas, used for preconcentrating the chlorophenols was prepared, stored and regenerated according to

F. Nauarro-Villoslada

240 &P

et al. /Analytica

Chimica Acta 308 (1995) 238-245

1

q(ml/min)

Fig. 1. Schematic

Edgerton et al. [7]. HPLC-grade Erba was used in the elution. 2.3. Flow-injection

diagram of the flow system. P.S. = Phase separator,

methanol from Carlo

procedure

The FIA system used is shown in Fig. 1. The sample (250 ~1) was injected into a borate buffer carrier solution at 1.5 ml min- ‘. The carrier solution then joined the TBA solution, whose flow-rate was 1.2 ml min- I. The reaction to form the ion pairs took place in a 30 cm X 0.3 mm i.d. reactor coil. These were extracted with chloroform, which was introduced in the system by a displacement technique, and joined the carrier solution in a T-shaped segmenter. Extraction took place in a 130 cm X 0.3 mm i.d. reactor. Separation of the organic and aqueous phases was performed in the membrane phase separator. The spectra around the peak maximum of the organic phase for each injection was collected every 2 nm over the wavelength range 200-430 nm every 0.6 s with an integration time of 0.3 s. All solutions were measured five times and the overall mean of the first derivative spectra of the five injections was stored and used for all subsequent data processing. The calibration set consisted of 50 random mixtures containing chlorophenols in the concentration range O-10 mg 1-l at six levels. Table 1 shows the composition of the calibration samples. A further 35 samples were prepared and analysed as an independent test set. Home-made programs of the three multivariate methods were applied for simultaneous de-

W = waste.

termination of each chlorophenol. These were CLS, PLS and a Kalman filter. 2.4. Sample preconcentration

methods

procedure

Preconcentration was carried out by using a 60 X 5 mm i.d. column packed with 1.2 g of XAD-4 polymeric adsorbent. This column was saturated with methanol and conditioned with 3 M HCl at pH 2.0. A volume of 500 ml of water sample was spiked with chlorophenol in the concentration range O-200 ppb, filtered through a 0.45 Frn pore size nylon microfilter and 10 ml of 3 M HCl was added. The sample was drawn through the column at flow rate of 4 ml min ’ Chlorophenols were eluted with 10 ml of methanol at 2 ml min- ‘. Once the methanol was evaporated chlorophenols were dissolved in the borate buffer (10 ml) and injected into the FIA system.

3. Results and discussion 3.1. Selection of working conditions for the flow system Hydrodynamic and chemical variables affecting the flow system were studied. Optimum chemical variables were coincident with those established for the batch method [14] except the TBA concentration. Because of the loss of sensitivity compared with the batch method and the poor reproducibility of the re-

241

F. Nauarro-Villoslada et al. /Analytica Chimica Acta 308 (199.5) 238-245 Table 1 Composition Sample

of the calibration samples (mg I-‘) 2-Chlorophenol

4-Chloro-3-methylphenol

2,4-Dichlorophenol

2,4,6-Trichlorophenol

Pentachlorophenol

1

2.0

2.0

1.9

2.1

2.0

2

3.9

4.0

4.1

4.0

3.9

3

6.1

5.9

6.0

6.1

6.1

4

8.1

7.9

7.9

7.9

8.1

5

10.0

9.9

10.1

10.0

10.0

6

10.0

6.1

4.1

6.1

8.1

7

8.1

9.9

1.9

6.1

3.9

8

8.1

7.9

4.1

6.1

2.0

9

2.0

4.0

10.1

6.1

10.0

10

6.1

9.9

7.9

7.9

6.1

11

10.0

4.0

6.0

10.0

8.1

12

8.1

4.0

1.9

2.1

3.9

13

2.0

2.0

10.1

10.0

6.1

14

10.0

9.9

1.9

10.0

3.9

15

6.1

7.9

7.9

10.0

6.1

16

8.1

4.0

6.0

10.0

6.1

17

2.0

2.0

6.0

6.1

6.1

18

2.0

4.0

1.9

2.1

8.1

19

10.0

4.0

7.9

4.0

8.1

20

6.1

4.0

4.1

7.9

10.0

21

6.1

8.1

2.0

0.0

4.0

22

2.0

4.3

6.0

6.0

0.0

23

7.9

6.1

10.0

6.0

0.0 4.0

24

7.9

2.0

0.0

6.0

25

0.0

2.0

8.0

10.2

6.0

26

0.0

6.1

10.0

2.0

4.0

27

2.0

9.9

2.0

0.0

8.0

28

10.2

0.0

8.0

4.0

2.0

29

4.1

6.1

0.0

10.2

2.0

30

7.9

0.0

2.0

4.0

6.0

31

0.0

2.0

6.0

0.0

10.0

32

4.1

8.1

6.0

0.0

0.0

33

0.0

0.0

10.0

8.2

4.0

34

7.9

9.9

0.0

0.0

6.0

35

0.0

2.0

6.0

4.0

0.0

36

4.1

6.1

0.0

8.2

0.0

37

0.0

6.1

0.0

8.2

2.0

38

2.0

0.0

4.0

0.0

6.0

39

4.1

0.0

0.0

6.0

4.0

40

7.9

0.0

4.0

6.0

0.0

41

2.0

0.0

0.0

4.0

0.0

42

0.0

4.3

0.0

0.0

2.0

43

2.0

0.0

0.0

0.0

6.0

43

0.0

2.0

0.0

8.2

0.0

45

0.0

9.9

4.0

0.0

0.0

46

0.0

0.0

10.0

0.0

4.0

47

0.0

0.0

6.0

4.0

0.0

48

2.0

9.9

0.0

0.0

0.0

49

0.0

0.0

0.0

6.0

2.0

50

4.1

0.0

6.0

0.0

0.1,

242 Table 2 Results of optimization

F. Nauarro-Villoslada

of working conditions

Variable

Injection volume ( ~1) Flow rate (ml min-’ ): Carrier TBA CHCI, Reactor: Length (cm) i.d. (mm) Extraction coil: Length (cm) i.d. (mm) Chemical variables: TBA nitrate (mg I-’ ) Ionic strength (NaCI, M)

Range studied 50-500

et al./Analytica

for the flow system Value selected

250

0.8-2.0 0.8-1.5 0.6-1.2

1.5 1.2 0.6

10-30 0.3-0.7

30 0.3

30-180 0.3-0.7

130 0.3

100-1800 0.00-0.09

1500 0.05

Chimica Acta 308 (1995) 238-245

of the five chlorophenols. Table 2 shows the range and variables tested and the values selected for 10 mg l- ’ of chlorophenol. An injection volume of 250 ~1 was used since the analytical signal was not significantly increased by larger volumes. The flow-rates used provided the best sensitivity and reproducibility. The 130 cm X 0.3 mm i.d. extraction coil yielded the optimum mass transfer. A PTFE phase separator containing a 8.5 ~1 mini-chamber and a 3.0 pm pore size Fluoropore membrane was used; it led to accurate and reproducible results with low V,/V, ratios (ca. l/4). It was advisable to pump the waste out of the flow cell as this drastically improved the reproducibility. Fig. 2 shows the highly overlapping of Table 3 Condition

sults for 2,4,6-trichlorophenol, the TBA concentration and ionic strength were studied in the flow system. The values selected for these variables were the most appropriate for the simultaneous determination

Wavelength

number for different wavelength range (nm)

250-380 250-320 280-360 280-340 282-336 286-33X 286-350 280-330 280-380 270-338 280-310 270-380

Table 4 Relative error of prediction

ranges

Cond(K) 12.0 11.2 6.7 5.5 5.0 4.4 5.6 5.9 6.7 10.4 10.7 11.2

of the test set (%*o)

Phenol

CLS

Kalman filter

PLS

2-Chlorophenol 4-Chloro-3-methylphenol 2,4-Dichlorophenol 2,4.6-Trichlorophenol Pentachlorophenol MREP (%)

7.0 7.5 9.3 11.8 9.7 8.4

6.0 9.2 9.5 9.3 9.5 8.3

8.1 9.9 8.5 16.6 9.4 9.9

MREP = Mean relative error of prediction. Table 5 Standard error of prediction

Fig. 2. Spectra of the five chlorophenols; (b) first-derivative, each at 1 mg 1-l.

(a) direct absorbance

and

(SEP) of the test set (mg 1-l )

Phenol

CLS

Kalman filter

PLS

2-Chlorophenol 4-Chloro-3-methylphenol 2,4-Dichlorophenol 2,4,6-Trichlorophenol Pentachlorophenol

0.4 0.4 0.5 0.6 0.5

0.3 0.5 0.5 0.4 0.5

0.5 0.5 0.4 0.8 0.5

F. Nauarro-Villoslada Table 6 Recoveries

et al. /Analytica

243

Chimica Acta 308 (1995) 238-245

obtained from purified water spiked with the chlorophenols Added a

Phenol

(ILg 1-l)

Kalman filter

CLS Mean recovery

R.S.D. (%)

(%o) 2-Chlorophenol 4-Chloro-3-methylphenol 2,4-Dichlorophenol 2,4,6-Trichlorophenol Pentachlorophenol

120 120 100 140 120

Mean recovery

PLS R.S.D. (%)

95 98 98 91 98

2 3 3 4 4

98 108 96 103 98

Mean recovery

R.S.D. (%I

(%o)

(%I 2 4 3 4 4

95 93 104 87 97

1 1 4 8 4

a Mean of 5 determinations.

chlorophenol spectra and their first-derivative under the working conditions selected. 3.2. Determination

of chlorophenols

spectra

in test samples

Firstly, the optimal wavelength range for achieving acceptable accuracy and precision in the multicomponent analysis was chosen. Error propagation is important in establishing the optimal wavelength range for quantitation. This error can be represented by the condition number of the calibration matrix, cond(K) [28]; a large condition number represents a large error in estimating the analyte concentration as well as a high degree of non-orthogonality in the spectra. Instead of selecting a wavelength range with the best sensitivity, it must be selected with the smallest error amplification, i.e., smallest cond(K) [27]. So this criterion was used in the three methods studied. The cond(K) value was calculated for a set of different wavelength ranges between 250-380 nm, which is the interval where the chlorophenols present absorbance. Table 3 shows the calculated cond(K) values for several wavelength ranges. The lowest cond(K) value was 4.4 in the wavelength range 286-338 nm and, therefore, was selected for subsequent calculations. The selection of the number of factors or latent variables for PLS was made by the cross-validation method using different groups of samples and criteria of goodness of fit [25]. The number of factors selected was that which agreed with several of the criteria. This number of factor was 6 using meancentered data. An outlier analysis was made taking into account the leverage and the spectral residuals via an F-test at the 95% confidence level [29]. Samples with high leverage (maximum leverage > 2(1 + factor

number)/No. of samples) and an F-value larger than a certain percentile in the F-distribution were considered as outliers. In this case several samples showed a large F-value but not a high leverage so no samples were removed. The analytical precision of the different calibration methods was evaluated as the relative error of prediction (REP). The results (Table 4) show that the best REP for each chlorophenol depends of the method used but, in general, the best results were obtained with the Kalman filter (REP < 10%). On the other hand, the standard error of prediction (SEP) was used to evaluate if there are significant differences between the concentrations found for each chlorophenol and method via an F-test at the 95% confidence level. Table 5 shows the value of the SEP for each chlorophenol and method. The F-test showed no significant differences for 2-chlorophenol, 4-chloro-3-methylphenol, 2,4-dichlorophenol and pentachlorophenol, but there were significant differences for 2,4,6-trichlorophenol between the PLS and Kalman filter procedures, the latter giving better results. 3.3. Preconcentration water

of chlorophenols

in purified

Water samples (500 ml) spiked with chlorophenol in the concentration range O-200 ng mll ’ were preconcentrated on XAD-4 resin. The preconcentration study was carried out in purified water to evaluate recoveries of each chlorophenol. As shown in Table 6 results of chlorophenol recoveries are dependent upon the multivariate calibration method applied. Recoveries averaged better than 95% for the CLS and Kalman filter methods and better than 87% for the PLS method.

244

F. Nac’arro-Villoslada

et al./Analytica

Chimica Acta 308 (1995) 238-245

Table 7 Results obtained from tap water spiked with chlorophenols Sample

2-Chlorophenol 4-Chloro-3-methylphenol 2,4-Dichlorophenol 2,4,6-Trichlorophenol Pentachlorophenol 2-Chlorophenol 4-Chloro-3-methylphenol 2,4-Dichlorophenol 2,4,6-Trichlorophenol Pentachlorophenol 2-Chlorophenol 4-Chloro-3-methylphenol 2,4-Dichlorophenol 2,4,6-Trichlorophenol Pentachlorophenol 2-Chlorophenol 4-Chloro-3-methylphenol 2,4-Dichlorophenol 2,4,6-Trichlorophenol Pentachlorophenol 2-Chlorophenol 4-Chloro-3-methylphenol 2,4-Dichlorophenol 2,4,6-Trichlorophenol Pentachlorophenol 2-Chlorophenol 4-Chloro-3-methylphenol 2,4-Dichlorophenol 2,4,6-Trichlorophenol Pentachlorophenol 2-Chlorophenol 4-Chloro-3-methylphenol 2,4-Dichlorophenol 2,4,6-Trichlorophenol Pentachlorophenol 2-Chlorophenol 4-Chloro-3-methylphenol 2,4-Dichlorophenol 2,4,6-Trichlorophenol Pentachlorophenol 2-Chlorophenol 4-Chloro-3-methylphenol 2,4-Dichlorophenol 2,4,6-Trichlorophenol Pentachlorophenol

3.4. Determination water samples

Added (pg

I-‘)

140 120 120 100 100 100 80 140 120 80 140 140 80 120 120 200 180 140 140 60 0 160 160 140 140 60 0 120 80 120 180 60 0 160 60 160 100 100 0 140 160 180 60 60 0

of chlorophenols

CLS

Kalman filter

Found ( pg I-‘)

Error (%I

Found (pg

138 122 128 102 106 96 78 140 122 76 134 134 76 118 126 194 178 144 132 58

1.4 1.7 6.7 2.0 6.0 4.0 2.5 0.0 1.7 5.0 4.3 4.3 5.0 1.7 5.0 3.0 1.1 2.9 5.7 3.3 _

166 162 158 148 62

3.7 1.2 13 5.7 3.3 _

12; 94 124 188 56 0 170 66 158 96 106 0 140 1.50 166 58 56 0

1.7 17 3.3 4.4 6.7 _

142 134 128 112 106 98 86 138 128 76 138 146 74 126 128 200 196 146 144 58 2 178 158 170 148 64 6 120 98 124 192 66 0 174 66 162 106 104 0 140 156 182 58 66 0

added to tap

The proposed method was applied to the determination of chlorophenols added to tap water samples.

6.2 10 1.2 4.0 6.0 _ 0.0 6.2 7.8 3.3 6.7 _

1-l)

PLS Error (%D) Found ( Kg l- ‘)

Error (%)

1.4 12 6.7 12 6.0 2.0 7.5 1.4 6.7 5.0 1.4 4.3 7.5 5.0 6.7 0.0 8.9 4.3 2.9 3.3 _

1.4 3.3 13 12 4.0 0.0 10 8.6 18 12 2.9 10 7.5 18 1.7 3.0 11 17 33 13 _

11 1.2 21 5.7 6.7 _ 0.0 22 3.3 6.7 10 _ 8.7 10 1.2 6.0 4.0 _ 0.0 2.5 1.1 3.3 10 -

138 116 136 88 104 100 72 152 98 70 136 126 86 98 122 194 160 164 94 52 0 152 174 134 144 68 6 130 82 120 192 58 2 158 62 160 98 104 0 140 154 164 66 42 0

5.0 8.7 4.3 2.9 13 _ 8.3 2.5 0.0 6.7 3.3 _ 1.2 3.3 0.0 2.0 4.0 _ 0.0 3.7 8.9 10 30 _

Table 7 shows the results. The three multivariate calibration methods gave acceptable errors (ca. 5%) for each chlorophenol except for 2,4,6-trichlorophenol with errors around 15%. It must be emphasized that the errors for 2,4,6-trichlorophenol and pen-

F. Naoarro-Villoslada Table 8 Standard error of prediction

et al. /Analytica

(SEP) for spiked tap water ( fig l-

Phenol

CLS

Kalman filter

PLS

2Chlorophenol 4-Chloro-3-methylphenol 2,4-Dichlorophenol 2,4,6-Trichlorophenol Pentachlorophenol

5 6 4 9 5

5 10 4 14 5

6 11 12 20 5

’)

tachlorophenol are quite reasonable in spite of the completely overlap of their spectra (Fig. 2). On the other hand, it was possible to determine 2,4,6-trichlorophenol with good precision in the FIA system while it was not possible in the batch method, likely because of the instability of its ion-pair with TBA. This is because, in the FIA system, the spectra are measured at a very reproducible fixed time. Taking into account the SEP (Table 8) no significant differences for 2-chlorophenol, 4-chloro-3methylphenol, 2,4-dichlorophenol and pentachlorophenol were observed, but for 2,4,6-trichlorophenol differences between the PLS and CLS methods and between the PLS and Kalman filter methods were observed. The best results were obtained using the CLS method (SEP < 9 pg l- ‘1.

4. Conclusions The proposed flow-injection method is suitable for simultaneous determination of five priority pollutant chlorophenols added to tap water at the pg 1-r level with previous preconcentration. The comparative studies show the best results expressed as SEP for the CLS method. The FIA system made possible the determination of 2,4,6-trichlorophenol which yields an ion pair that is not very stable.

Acknowledgements The financial support of the Spanish DGICYT, project PB92-0192, is gratefully acknowledged.

References [l] P. Zhang and D. Littlejohn,

Analyst,

118 (1993) 1065.

Chimica Acta 308 (1995) 238-245

245

[2] W. Frenzel, J. Oleksy-Frenzel and J. Moller, Anal. Chim. Acta, 261 (1992) 253. [3] M.K. Fayyad, M.A. Alawi and T.J. El-Ahmed, Chromatographia, 28 (19891 46.5. [4] P.J.M. Kwakman, D.A. Kamminga, U.A.Th. Brinkman and G.J. De Jong, J. Chromatogr., 553 (19911 345. [5] E.C.V. Butler and G. Dal Pont, J. Chromatogr., 609 (1992) 113. [6] R.G. Melcher, D.W. Bakke and G.H Hughes, Anal. Chem., 64 (1992) 2258. [7] T.R. Edgerton, R.F. Moseman, E.M. Lores and L.H. Wright, Anal. Chem., 52 (19801 1774. [8] G.W. Patton, L.L. McConnell, M.T. Zaranski and T.F. Bidleman, Anal. Chem., 64 (19921 2858. [9] H. Kontsas, C. Rosenberg, P. Jappinen and M.L. Riekkola, J. Chromatogr., 636 (1993) 255. [lo] A. Morales, D.A. Birkholz and S.E. Hrudey, Water Environ. Res., 64 (19921 669. [ll] Ch.P. Ong, H.K. Lee and S.F.Y. Li, .I. Chromatogr. Sci., 30 (1992) 319. [12] H.B. Lee, T.E. Peart and R.L. Hong-You, J. Chromatogr., 605 (1992) 109. [13] P. Praus and V. Dombek, Anal. Chim. Acta, 277 (1993) 97. [14] F. Navarro-Villoslada, M.E. Leon-Gonzalez, L.V. PCrezArribas, M.J. Santos-Delgado and L.M. Polo-Diez, Microchem. J., 44 (19911 339. [15] A. Cladera, E. Gbmez, J.M. Estela and V. Cerda, Anal. Chim. Acta, 267 (1992) 95. [16] A. Cladera, E. Gbmez, J.M. Estela and V. Cerda, Anal. Chim. Acta, 267 (1992) 103. [17] V. Kuban, CRC, Crit. Rev. Anal. Chem., 23 (1992) 15. [18] C. Ridder and L. Norgaard, Chemom. Intell. Lab. Syst., 14 (19921 297. [19] M. Blanco, J. Coello, H. Iturriaga, S. Maspoch. M. Redon and J. Riba, Anal. Chim. Acta, 259 (1992) 219. [20] P. MacLaurin, P. Worsfold, P. Norman and M. Crane, Analyst, 118 (19931 617. [21] D.A. Whitman, M.B. Seasholtz, G.D. Christian, J. Ruzicka and B.R. Kowalski, Anal. Chem., 63 (1991) 775. [22] A. Cladera, E. G6mez, J.M. Estela, V. Cerda, A. Alvarez Ossorio, F. Rincon and F. Salva, Int. J. Environ. Anal. Chem., 45 (1991) 143. [23] E.V. Thomas and D.M Haaland, Anal. Chem., 62 (1990) 1091. [24] P.M. Lang and J.H. Kalivas, J. Chemom., 7 (1993) 153. [25] B.R. Kowalski and M.B. Seasholtz, J. Chemom., 5 (1991) 129. [26] S.C. Rutan, J. Chemom., 4 (1990) 103. [27] L.V. Perez-Arribas, F. Navarro-Villoslada, M.E. LeonGonzalez and L.M. Polo-Diez, J. Chemom., 7 (1993) 267. [28] J.H. Kalivas and P.M. Lang, J. Chemom., 3 (1989) 443. [29] H. Martens and T. Naes, Multivariate Calibration, Wiley, Chichester, 1989, p. 274.