Assessment of uncertainty in pesticide multiresidue ...hera.ugr.es/doi/15085983.pdf · carrier gas used was helium (purity 99.999%). Mass spectrometer settings: solvent delay, 4.5min;

Analytica Chimica Acta 454 (2002) 297–314

Assessment of uncertainty in pesticide multiresidue analyticalmethods: main sources and estimation

L. Cuadros-Rodrıgueza, M.E. Hernández Torresb, E. Almansa Lópeza,F.J. Egea Gonzálezc, F.J. Arrebola Liébanasc, J.L. Martınez Vidalc,∗

a Department of Analytical Chemistry, School of Qualimetrics, University of Granada, Granada, Spainb Centro Universitario Analıtico Municipal (CUAM), El Ejido, Almerıa, Spainc Department of Analytical Chemistry, University of Almerıa, Almerıa, Spain

Received 26 June 2001; received in revised form 1 November 2001; accepted 15 November 2001

Abstract

The estimation of the uncertainty associated to analytical methods is necessary in order to establish the comparability ofresults. Multiresidue analytical methods lack very often of information about uncertainty of results with likely implicationswhen results are compared with maximum residue levels (MRL) established by regulations. An adequate identificationand estimation of each uncertainty source allows to laboratories to establish the accuracy of results and to balance withtime-consuming and costs. © 2002 Elsevier Science B.V. All rights reserved.

Keywords:Uncertainty; Gas chromatography; Pesticide analysis

1. Introduction

The analytical properties are classified among otherways [1] in capital, basic, and accessory. A capitalone is the accuracy [2], as a conjunction of the basicsprecision, trueness and inertia, related with such prop-erties we find meteorological properties, such as trace-ability and uncertainty. These are intimately linked,having into account the formal definitions of traceabil-ity and uncertainty [3], we could state that uncertaintycharacterises the strength of the links in the chain oftraceability and the agreement to be expected betweenmeasurements. Uncertainty takes into account eitherrandom as systematic errors and gives information

∗ Corresponding author. Tel.:+34-950015429;fax: +34-950015483.E-mail address:[email protected] (J.L. Martınez Vidal).

about the range in which a result can be expected.The error cannot be determined in most of analyticalmethods unless the true value is stated, but the uncer-tainty can be estimated from the analytical method,which should be described in a detailed operating pro-cedure. This point of view, addresses the uncertaintyas a probabilistic estimation of the maximum error ofa measurement.

Three approaches are already proposed: bottom-up[4,5], top-down [6] and in-house validation [7,8] forthe expression of uncertainty. The first approach con-siders the division of the analytical method into itssteps and the identification, quantification and com-bination of all uncertainty sources. The InternationalStandards Organisation (ISO) produced a guide forthe harmonisation of the expression of results with un-certainty, which was adapted by Eurachem to the ana-lytical problem. The second approach includes the use

0003-2670/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved.PII: S0003-2670(01)01546-X

298 L. Cuadros-Rodrıguez et al. / Analytica Chimica Acta 454 (2002) 297–314

of interlaboratory information and the third approachinvolves the information obtained from repeatedanalyses derived from in-house validation of analyti-cal methods. In this context it seems sense for a givenpesticide residues laboratory, to use the bottom-upapproach in conjunction with in-house validation datafor estimating the uncertainty derived from each stepof the analytical method [9–11].

All this information should be considered in theframe of the analytical problem: pesticide residueanalysis in vegetables has the lack in certified ref-erence materials, CRMs, consider a wide range ofanalytes, which should be determined at very differ-ent concentration levels established by the maximumresidue levels (MRLs), and finally there is a widerange of commodities with different matrix effect inthe determination of the analytes.

On other hand, the definition of uncertainty [3], in-dicates that results should be given without systematicerrors. In pesticide residue analysis, the situation isthat recovery factors of most of analytes are differentthan 100%, so that, in a first consideration, a correctionof this recovery should be necessary for the reliabilityof results [12]. Nevertheless, having into account againthe diversity of concentration ranges and the largenumber of analytes, it make difficult to estimate a cor-rection factor for each combination, adopting the ac-cepted criteria that no correction is better than a wrongcorrection. The European Commission [13] consid-ers acceptable recovery factors within 70–110%,supporting the establishment of MRLs on thisbasis.

However a couple of questions remain: the first oneis the case in which a pesticide is found in a sampleat a concentration close to its MRL or even slightlygreater. In a practical sense, if such limit is in the con-centration range defined by the measurement and itsuncertainty, the result should be considered as nega-tive, in fact, only when the MRL is minor than theamount found minus its uncertainty, the result shouldbe stated as positive. The other question is the caseof results obtained in different laboratories, the un-certainty can led to consider them as concordant ordiscrepant depending on the inclusion of the allowedlimit in the intervals given by the uncertainty of resultsin each laboratory.

So that, should uncertainties be considered in thelegislation for a better comparison of results? Is the

above statement about the correction of recovery fac-tors scientifically supported? Having into account thatmethods should be validated for a range of analytesin a range of representative commodities, allowing toestimate recovery factors for each analyte and eachcommodity.

This paper presents a methodology for estimatingthe uncertainty associated to a multiresidue analyticalmethod in cucumber matrix, through the bottom-upapproach and on the basis of in-house validation data.All data appearing in this paper were obtained with themultiresidue method (MRM) proposed, which meetcompliance with EN45001 requirements, since it wasimplemented in a pesticide residue analysis laboratoryas a routine method and accredited. The uncertaintyof each step is estimated identifying which of themare relevant in the global uncertainty, finally the sig-nificance of recovery factors is statistically calculatedincluding the estimation of the results with and with-out correction.

2. Experimental

2.1. Chemicals

The solvents used weren-hexane, acetone,cyclohexane and dichloromethane (residue analysisgrade, Panreac, Barcelona, Spain). Anhydrous sodiumsulphate for residue analysis was purchased from Pan-reac. All pesticide standard reference materials wereobtained from Dr. Ehrenstorfer (Augsburg, Germany).The purity of these standard pesticides is given inTable 1. The following pesticides were tested.

(a) Pesticides determined by GC–NPD: Dichlorvos,methamidofos, acephate, heptenophos, ethopro-phos, dimethoate, diazinon, etrimfos, parathion-methyl, chlorpyrifos-methyl, pirimiphos-methyl,malathion, parathion-ethyl, chlorpyrifos, chlorfen-vinphos, triadimenol, fenamiphos, myclobutanil,buprofezin, cyproconazole, ethion and carbophe-nothion.

(b) Pesticides determined by GC–ECD: �-HCH,chlorothalonil, vinclozolin, dichlofluanid, triadime-fon, chlozolinate, procymidone, endosulfan-�,endosulfan-�, endosulfan-sulphato, nuarimol, ipro-dione, bromopropylate, tetradifon and acrinathrin.

L. Cuadros-Rodrıguez et al. / Analytica Chimica Acta 454 (2002) 297–314 299

Table 1Relevant data for calculating the uncertainty associated to standardpreparation

Pesticide Purity(%)

Tolerance(%)

Mass (m)a

(mg)VP1

(�l)b

ECD�-HCH 99.4 0.1 13.0 1940Chlorothalonil 97.0 0.1 10.7 2405Vinclozolin 97.6 0.1 10.4 2465Dichlofluanid 98.0 0.5 11.1 4760Triadimefon 98.6 0.1 10.2 500Chlozolinate 98.0 0.1 10.4 495Procymidone 99.7 0.5 13.5 1860Endosulfan-� 98.3 0.1 9.1 2795Endosulfan-� 99.1 0.1 10.3 2450Endosulfan-s 98.1 0.1 12.2 2090Nuarimol 99.2 0.5 10.1 500Iprodione 99.5 0.1 10.5 2405Bromopropylate 97.4 0.1 10.3 2490Tetradifon 99.9 0.1 9.4 2660Acrinathrin 99.9 0.1 3.0 167

NPDDichlorvos 99.0 0.1 17.0 1180Methamidofos 94.0 0.1 11.6 305Acephate 98.9 0.1 11.1 910Heptenophos 96.4 0.1 10.4 995Ethoprophos 92.0 0.1 10.9 500Dimetoate 98.7 0.1 11.5 1760Diazinon 94.0 0.1 10.6 1005Parathion-methyl 98.6 0.1 11.3 1795Chlorpyrifos-methyl 98.8 0.1 13.8 1535Pirimiphos-methyl 98.0 0.1 10.7 1905Malathion 98.7 0.1 10.8 1510Parathion-ethyl 99.0 0.1 10.1 2000Chlorpyrifos 98.7 0.1 11.4 1835Chlorfenvinphos 94.0 0.1 10.3 1895Triadimenol 98.0 0.1 10.2 2000Fenamiphos 97.1 0.1 10.0 2060Myclobutanil 99.6 0.1 10.0 480Buprofezin 99.4 0.1 10.0 830Cyproconazole 99.5 0.1 11.0 495Ethion 94.0 0.1 10.7 1990Carbophenothion 96.0 0.1 10.4 500

a The volume of the primary standard solutions was 50 ml inall cases excepting achrinathrin (10 ml).

b VP1 is the volume of primary standard solution measuredfor preparing the secondary standard solution. The final volumeof this solution was 50 ml.

2.2. Equipment

Three gas chromatographs were used: a Perkin-Elmermodel 8500 equipped with an electron capture de-tector (ECD,63Ni); a Fisons model 8000 equipped

with an nitrogen-phosphorus detector (NPD) and aninjector AS Fisons and a Saturn 2000 ion trap massspectrometer from Varian Instruments (Sunnyvale,CA, USA) equipped with an autosampler 8200, anda split/splitless programmed temperature injectorSPI/1078 operated in the splitless mode.

GC–ECD operating conditions: injector temper-ature, 250◦C; detector temperature, 350◦C; initialoven temperature, 180◦C for 5 min, raised at 3◦C/minto 250◦C, and then held at 250◦C for 2 min. Thecarrier gas was nitrogen at 10 ml/min. A fused sil-ica semicapillary (HP-1) column containing 100%methylpolysiloxane as stationary phase (25 m length,0.53 mm internal diameter (i.d.) and 1.0 mm filmthickness) was used for the separation.

GC–NPD operating conditions: injector tempera-ture, 250◦C; detector temperature, 300◦C; splitlesstime, 2 min; initial oven temperature, 90◦C for 1 min,raised at 15◦C/min to 170◦C for 0 min, raised at2◦C/min to 220◦C for 0 min, raised at 10◦C/min to255◦C, and then held at 255◦C for 5 min. The carriergas was nitrogen at 1 ml/min. A fused silica capillary(HP-1) column containing 100% methylpolysiloxaneas stationary phase (60 m length, 0.25 mm i.d. and0.25�m film thickness) was used for the separationin the GC.

GC–MS operating conditions were: initial columntemperature, 60◦C (2.9 min), increased at 40◦C/minto 150◦C and finally increased at 5◦C/min to 275◦C(held for 10 min); initial injector temperature, 60◦C(0.3 min) and increased at 100◦C/min to 280◦C (held30 min); manifold, transfer-line and trap temperatureswere 45, 260 and 200◦C, respectively; flow-rate, 1�l/sand injection volume, 5�l. A DB5-MS column (30 m,0.25 mm i.d. and 0.25�m film thickness) was em-ployed. The ion trap mass spectrometer was operatedin the electron ionisation (EI) mode and the MS/MSoption was used. The computer, which controlled thesystem, had an EI–MS/MS library specially createdfor the target analytes in our experimental conditions.In addition, other EI–MS libraries were available. Thecarrier gas used was helium (purity 99.999%).

Mass spectrometer settings: solvent delay, 4.5 min;70 eV of electron impact energy and scan rate,0.6 scans/s.

For GC–MS/MS, the sample was injected underthe gas chromatographic conditions described forGC–MS.


2.3. Analytical procedures

2.3.1. Standards preparationA stock solution of each pesticide (primary stan-

dard solutions) was prepared dissolving pure standardin acetone (for pesticides analysed by GC–NPD) andin n-hexane (for pesticides analysed by GC–ECD).Working standard solutions containing a mixtureof the analytes (secondary standard solutions) were

Table 2Data calibration and parameters calibrationsa

Compound Concentration (mg/l) b a sb sresid Repeatability(%)

Intermediateprecision (%)

R (%)

ECD�-HCH 0.250–0.500–1.000 7.250 −0.405 0.121 0.146 1.9 4.8 81.8Chlorothalonil 0.013–0.025–0.050 3.322 −0.145 0.041 0.050 2.5 6.4 76.9Vinclozolin 0.250–0.500–1.000 3.021 −0.110 0.032 0.039 1.8 5.7 83.0Dichlofluanid 0.500–1.000–2.000 3.302 −0.367 0.028 0.068 1.7 4.5 76.1Triadimefon 0.050–0.100–0.200 1.549 −0.027 0.029 0.007 4.2 8.3 85.0Chlozolinate 0.050–0.100–0.200 2.907 −0.006 0.038 0.009 2.5 5.2 82.5Procymidone 0.250–0.500–1.000 0.377 −0.009 0.008 0.009 1.2 5.9 88.0Endosulfan-� 0.250–0.500–1.000 5.010 −0.133 0.033 0.039 1.8 5.3 88.8Endosulfan-� 0.250–0.500–1.000 3.321 −0.151 0.036 0.044 2.9 4.4 86.3Endosulfan-s 0.250–0.500–1.000 2.248 −0.105 0.026 0.031 1.6 5.4 87.6Nuarimol 0.050–0.100–0.200 1.640 0.015 0.030 0.007 3.2 6.5 101.3Iprodione 0.250–0.500–1.000 0.333 −0.009 0.008 0.009 2.6 6.9 87.9Bromopropylate 0.250–0.500–1.000 1.834 −0.007 0.021 0.026 2.0 4.9 90.0Tetradifon 0.250–0.500–1.000 1.839 0.019 0.028 0.034 2.8 5.6 90.5Acrinathrin 0.025–0.050–0.100 1.224 0.003 0.010 0.004 12.3 19.2 101.3

NPDDichlorvos 0.200–0.600–1.000 19.307 −0.294 0.586 0.741 3.2 6.0 89.6Methamidofos 0.050–0.150–0.350 1.155 0.042 0.047 0.015 12.0 19.6 95.0Acephate 0.100–0.300–0.700 1.712 −0.012 0.028 0.018 4.9 12.5 88.3Heptenophos 0.100–0.300–0.700 2.719 0.026 0.053 0.033 3.6 13.0 87.5Ethoprophos 0.050–0.150–0.350 13.799 0.041 0.176 0.056 10.9 12.5 90.0Dimetoate 0.200–0.600–1.400 2.905 0.010 0.092 0.116 3.2 8.9 88.8Diazinon 0.100–0.300–0.700 5.731 −0.206 0.116 0.074 11.2 13.7 89.2Parathion-methyl 0.200–0.600–1.400 2.037 0.160 0.062 0.078 4.8 9.0 86.3Chlorpyrifos-methyl 0.200–0.600–1.400 3.485 0.047 0.117 0.148 3.7 11.8 84.6Pirimiphos-methyl 0.200–0.600–1.400 2.376 0.199 0.094 0.064 3.2 20.4 89.6Malathion 0.200–0.600–1.400 1.537 0.104 0.073 0.093 4.2 13.1 87.9Parathion-ethyl 0.200–0.600–1.400 2.088 0.168 0.103 0.130 3.2 7.9 85.0Chlorpyrifos 0.200–0.600–1.400 2.112 0.160 0.091 0.115 5.0 16.9 91.3Chlorfenvinphos 0.200–0.600–1.400 2.161 −0.119 0.086 0.109 3.3 12.6 86.3Triadimenol 0.200–0.600–1.400 0.178 0.008 0.007 0.009 5.4 24.4 89.2Fenamiphos 0.200–0.600–1.400 0.816 0.036 0.035 0.044 2.2 18.8 88.3Myclobutanil 0.050–0.150–0.350 1.971 0.036 0.053 0.017 9.0 25.7 88.3Buprofezin 0.100–0.300–0.700 0.520 0.010 0.013 0.008 9.9 12.9 85.0Cyproconazole 0.050–0.150–0.350 1.170 0.014 0.022 0.007 9.6 14.0 90.0Ethion 0.200–0.600–1.400 2.702 0.769 0.145 0.184 3.4 13.3 88.3Carbofenothion 0.050–0.150–0.350 5.155 0.073 0.115 0.036 7.8 13.4 81.7

a Slope, ‘b’; standard deviation of slope,sb; intercept, ‘a’ and standard deviation of residues,sresid; recovery factors (R); validationdata (repeatability and intermediate precision).

prepared from the above by appropriate solvent dilu-tions, using automatic pipettes and glass volumetricflasks (A class). They were stored in a refrigerator at4◦C. Table 1 shows the amount of each solid standardtaken for preparing the primary standard solution,and the volumes taken for preparing the secondarystandard solutions. Finally, calibration curves wereprepared at the concentration ranges given in Table 2,pippeting 50, 100, and 200�l of the secondary


Fig. 1. Schemes of the standards preparation (A) and extraction procedure (B).

standard solution for ECD analysis and 50, 150 and250�l for NPD analysis, and diluting them to 2 ml, ina volumetric flask with blank matrix extract (Fig. 1A).

2.3.2. Extraction and analysisThe extracting method used was similar to that used

by Martınez Vidal and co-workers [14], which con-sists in mixing 50 g of a chopped sample with 105 mlof dichloromethane, the mixture is homogenised withPolytron at 10 000 rpm for 2 min, then 100 g of an-hydrous sodium sulphate are added resting for 2 minand the mixture is filtered through a filter paper into a250 ml round-bottom flask and the cake was washedtwice with 20 ml of dichloromethane each time. Thesolvent is removed under vacuum at 40◦C in a rotaryevaporator until almost dry and then just to the point ofdryness with a slight N2 stream, being dissolved with5 ml of cyclohexane. For ECD analysis, 1 ml of thesample extract is diluted adding 3 ml ofn-hexane. In

all these steps, appropriate glass pipettes, A class, areused for measuring the volumes. The 40�l of internalstandard (IS) solution (10 mg/l of dieldrin in hexane)is added in a 2 ml volumetric flask, the volume madeup to 2 ml with the above solution and injected into theGC–ECD (1�l). For NPD analysis, 100�l of the IS(20 mg/l of caffeine in acetone) is added in a 2 ml vol-umetric flask and the volume made up to 2 ml with thesample extract and injected into the GC–NPD (1�l).For the MS confirmation, 1 ml of the NPD solution isdiluted to 2 ml with cyclohexane and injected into theGC–MS (5�l) (Fig. 1B).

2.3.3. Recovery studyThe recovery study was carried out spiking 50 g

of cucumber sample, which had not been treatedwith the pesticides, with a mixture of working stan-dard solutions containing all pesticides at the secondconcentration level defined for the calibration curves


(Table 2). After evaporation of the solvent using anitrogen stream, the sample was mixed thoroughlyand homogenised for 2 min. Then the sample wasextracted and analysed. Ten replicates of each recov-ery assay and 10 blank samples of cucumber wereperformed. Intermediate precision of recovery factorswas estimated performing the recovery experimenteach week during 3 months.

3. Uncertainty estimation: theoretical aspects

Depending on the way of expressing uncertainty,we find standard uncertainty(u(x)), expressed as astandard deviation, andexpanded uncertainty(U(x))which is calculated from a combined standard uncer-tainty and a coverage factork.

In some cases, it is feasible to use relative uncer-tainties (in both uncertainties), which represent thevalue of the uncertainty normalised. It is obtained asthe quotient between the standard uncertaintyu(x) andthe value ofx:

Urel(x) = U(x)

xor urel(x) = u(x)

x

Uncertainty estimation is simple in principle, thesteps involved are as follows.

• Specify the measurand. It should be clearly writtenthe relationship between the measurand and the in-put quantities upon which it depends, such as mea-sured quantities, constants and calibration standardvalues.

• Identify uncertainty sources. Listing the possiblesources of uncertainty, usually specified in the abovestep.

• Quantify uncertainty components. Estimating theuncertainty component associated with each poten-tial source of uncertainty identified. The differentcontributions to the overall uncertainty have to beexpressed as standard deviation which can be cal-culated depending on the data available:◦ from a standard deviation value: this value is di-

rectly used;◦ from a variation coefficient;◦ from the standard deviation of experimental data

sets;◦ from a declared uncertainty value, which is given

in a certificate of calibration;

◦ from a confidence interval;◦ from a maximum interval of variability;◦ from a range of limits(upper and lower limits);◦ finally from a given error value.

• Calculate combined uncertainty. The different con-tributions to the overall uncertainty have to be com-bined according to the appropriate rules for givinga combined standard uncertainty:

u(f ) =√

c2(x)u2(x) + c2(y)u2(y) + · · ·,f = f (x, y, . . . )

wherec is a sensibility coefficient associated to eachone of variables, given by the partial derivative ofthe function:c(x) = ∂f/∂x.

Applying the appropriate coverage factor, the ex-panded uncertainty will be obtained.

4. Results and discussion

4.1. Analysis

Chromatographic conditions were optimised forachieving a good resolution of the target pesticides.Figs. 2 and 3 show the ECD and NPD chromatogramsof a cucumber extract containing all the analytesand the IS, using the final selected chromatographicconditions. Table 3 summarises the retention timewindow (RTW) determined for all the compounds.The RTW is defined for each pesticide as the averageof the retention times, obtained from 10 replicates,plus or minus three times the S.D. of retention times(RT). Confirmation was carried out by GC–MS/MSanalysis, using the quantification ions summarised inTable 3. Target analytes were searched at RTW andwere identified by comparing their spectra with thoseEI–MS/MS libraries, stating a minimum spectral fitof 700 as confirmation requirement.

4.2. Validation

Calibration curves were obtained from matrix-matching calibration solutions using IS calibration.The lowest concentration level in the calibration curveis established as a practical determination limit (PDL),


Fig. 2. ECD chromatogram ofn-hexane extract of cucumber spiked at second concentration level of calibration curve and containing theIS (17.97 min).

which is defined as a percentage of the maximumresidue level (MRL) stated for each pesticide by theEuropean Union regulations in vegetables commodi-ties. All compounds exhibited good linearity in thestudied range. Determination coefficients (the squareof the correlation coefficients) found were higher than0.98 in all cases.

Limits of detection (LOD) and limits of quantifi-cation (LOQ) were calculated as 3- and 10-fold, re-spectively, the standard deviation of 10 blank samples,

containing the IS, divided by the slope of the calibra-tion curve. Table 3 summarises the LOD and LOQobtained for each pesticide. The values obtained arelower than their respective MRLs.

4.2.1. PrecisionThe repeatability of the method was tested by

determining the R.S.D. of chromatographic signalsobtained from a spiked sample analysed 10 times.The values found (Table 2) were lower using ECD,


Fig. 3. NPD chromatogram of an acetone extract of cucumber spiked at second concentration level of calibration curve and containing theIS (22.11 min).

1–5% (except achrinathrin with a 12.3% R.S.D.) thanNPD, which showed R.S.D. ranging between 2 and12%. The intermediate precision was determined bymeasuring the standard deviation of a set of spikedsamples, which were extracted and analysed (onereplicate) each week during 3 months (Table 2). Thisexperiment included in the R.S.D., the likely variationproduced by changes in solvent batches, analysts, ma-terial, and the usual chromatography maintenance. Asin the above case, values found expressed as %R.S.D.are higher for the compounds analysed by GC–NPD,being the maximum R.S.D. 25.7% for myclobutanyl.The compounds determined by GC–ECD presentvalues lower than 10% except for acrinathrin (19.2%).

4.3. Extraction procedure and recovery study

The extraction procedure described above was ap-plied to spiked cucumber samples for obtaining therecovery rates of pesticides. The extraction method isefficient for extracting pesticide residues from cucum-ber samples, as the analytes were determined with re-covery factors ranging between 75 and 101% (Table 2)at the second calibration level of spiked concentration,being the precision above explained (Table 2).

4.4. Quality control procedure

A quality control procedure was established for en-suring that results obtained are under statistical con-trol. This procedure consisted in incorporating to each

batch of samples a blank extract, a matrix-matchingcalibration solutions and three spiked samples.

Results were considered when the analysis of blankextracts showed that neither contamination nor degra-dation of sample had occurred, the recovery factors ofspiked samples were between 70 and 120% and thecalibration plots fit to lines with determination coeffi-cients higher than 0.95.

4.5. Uncertainty of results

The different aspects explained above for estimat-ing the standard uncertainties have been applied to themultiresidue analytical method. Tables 1 and 4 showthe relevant information for calculating uncertaintiesassociated to the preparation of primary standardsolutions and to the volumetric material, analyticalbalance and balance. Furthermore, calibration dataobtained daily during 1 month have been used foreach pesticide.

4.5.1. Identification of uncertainty sourcesThe analyte concentration in the sample, expressed

in mg/kg, is obtained from the equation:

CON = CA Fdil

CS

where CA is the analyte concentration obtained fromthe calibration (in mg/l);Fdil the dilution factor andCS the sample concentration in the extract (kg/l).


Table 3RTW, MS/MS data, LOD and LOQ

Compound Quantification ion (m/z) RTW LOD (mg/kg) LOQ (mg/kg)

ECD�-HCH 148 7.01–7.39 0.002 0.006Chlorothalonil 231 7.90–8.33 0.003 0.011Vinclozolin 115 10.06–10.48 0.001 0.005Dichlofluanid 123 11.53–11.99 0.002 0.006Triadimefon 144 12.78–13.24 0.006 0.021Chlozolinate 259 14.59–15.04 0.002 0.005Procymidone 255 15.13–15.61 0.007 0.023Endosulfan-� 170 16.27–16.78 0.001 0.004Endosulfan-� 170 18.92–19.46 0.002 0.006Endosulfan-s 237 21.27–21.83 0.001 0.004Nuarimol 139 22.62–23.21 0.002 0.007Iprodione 245 24.58–25.20 0.003 0.010Bromopropylate 181–187 25.32–25.81 0.001 0.010Tetradifon 197–203 26.26–26.85 0.001 0.019Acrinathrin 152 30.26–30.84 0.004 0.013

NPDDichlorvos 131 8.53–8.87 0.001 0.002Methamidofos 96 8.97–9.46 0.001 0.004Acephate 79 10.91–11.68 0.002 0.008Heptenophos 185 13.94–14–67 0.001 0.002Ethoprophos 94 15.49–16.15 0.001 0.012Dimetoate 79 17.77–18.36 0.001 0.010Diazinon 179 20.34–21.25 0.001 0.007Parathion-methyl 136 23.31–23.91 0.001 0.007Chlorpyrifos-methyl 208 24.17–25.16 0.001 0.010Pirimiphos-methyl 151 25.87–26.46 0.001 0.010Malathion 99 27.28–27.89 0.001 0.009Parathion-ethyl 142 28.21–28.76 0.001 0.005Chlorpyrifos 258 28.97–29.66 0.001 0.009Chlorfenvinphos 159 30.41–31.37 0.001 0.003Triadimenol 85 31.63–32.39 0.001 0.007Fenamiphos 195 33.03–33.72 0.002 0.005Myclobutanil 195 33.82–34.52 0.001 0.005Buprofezin 193 34.84–35.28 0.001 0.005Cyproconazole 125 35.44–36.03 0.004 0.012Ethion 175 36.28–36.83 0.001 0.005Carbophenothion 199 37.38–38.08 0.001 0.011

4.5.2. Quantification of standard uncertaintiesassociated to each step

The dispersion of results around the true value de-pends upon the following steps (Fig. 4).

1. Estimation of the analyte concentration from thecalibration curve.

2. Dilution factor of the sample extract.3. Calculation of the sample concentration.

The combined uncertainty (in terms of relative un-certainty) can be calculated with the expression:

urel(CON) =√

u2rel(CA) + u2

rel(Fdil) + u2rel(CS)

where each term of the sum, the relative standarduncertainty associated to each source identifiedabove.


Table 4Volumetric material used for preparing standards

Equipment Tolerancea (ml) Correctionb (�l) Variation coefficient (%) Uncertainty (k = 2)

Volumetric flask 50 ml ±0.05Volumetric flask 10 ml ±0.025Volumetric flask 2 ml ±0.025

Glass pipette 10 ml ±0.05Glass pipette 5 ml ±0.015Glass pipette 1 ml ±0.007

Micropipette 5–40�l 0.03 (5) ±2.000.02 (10) ±0.500.04 (40) ±0.50

Micropipette 40–200�l −0.18 (40) ±0.60−0.28 (70) ±0.21−0.60 (200) ±0.30

Micropipette 200–1000�l −0.32 (200) ±0.50−0.49 (300) ±0.33−1.47 (1000) ±0.30

Analytical balance (g) 0 ±0.001

Balance (g) ±0.033

a The tolerance is the confidence interval which is given by manufacturers.b The values in parenthesis are the volumes in which the correction value is established by manufacturers.

Fig. 4. Cause and effect diagram for the determination of pesticides in vegetable sample.


4.5.2.1. Estimation of the uncertainty derived from theestimation of the analyte concentration from the cal-ibration curve, u(CA). This is a combination of theuncertainties associated to the preparation of the cali-bration standard solutions (u2(std)), to the transforma-tion of the chromatographic signals in concentrations(u2(cal)) and to the repeatability of the measurements(u2(repet)). This combination is calculated as

u(CA) =√

u2(std) + u2(cal) + u2(repet)

4.5.2.1.1. Estimation of u2(std). It is estimatedfor each analyte, being a combination of the uncer-tainty derived from the preparation of the primaryand secondary standard solutions,u2(prim–secon),and from the preparation of the calibration curve atthree concentration levels by diluting the secondarystandard solution,u2(dil).

urel(std) =√

u2rel(prim–secon) + u2

rel(dil)

The concentration of the primary and secondarystandard solutions is given by the mass (m) of the solidstandard weighted in the analytical balance, the vol-ume (VF1) of the first dilution, in the case of the pri-mary standard; and by the volume (VP1) taken with a

pipette from the primary standard solution and the vol-ume (VF2) filled up in the second dilution in the caseof the secondary standard. Tables 1 and 4 show thedata used for the calculation of this term of the uncer-tainty. As an example data corresponding to�-HCHare included in the following steps:

Cprim–secon = m

VF1

VP1

VF2= 13.0 mg

50 ml

1.94 ml

50 ml

= 8.53× 10−3 mg/ml

so that the standard uncertainty associated to thesesteps can be obtained as

urel(Cprim–second)

=√

u2rel(m) + u2

rel(VF1) + u2rel(VP1) + u2

rel(VF2)

The uncertainty associated to the equipment whichhave been previously calibrated is calculated as:correction/

√3 + u. For example, the volume (VP1)

taken with a pipette was 1940�l and since the pipettewas used twice (1000�l + 940�l), the uncertaintyassociated is

u(VP1)1000 = correction√3

+ CV nominal

= −1.47√3

+ 0.003× 1000= 2.15

u(VP1)940 = correction√3

+ CV nominal

= −0.98√3

+ 0.0032× 940= 2.44

where the volume taken is different to the tabulatedvalue, the correction and the coefficient of variationare the mean value from the interval to which theybelong.

urel(Cprim–second)

=√

u2rel(m) + u2

rel(VF1) + u2rel(VP1) + u2

rel(VF2)

=

√√√√( 1/2

13.0

)2

+(

0.05/√

3

50

)2

+((

2.15

1000

)2

+(

2.44

940

)2)

+(

0.05/√

3

50

)2

= 0.039

In similar way, the uncertainty associated to thepreparation of the calibration curve is calculated foreach concentration level as

urel(dil) =√

u2rel(VP2) + u2

rel(VF3)

=

√√√√(0.006

100

)2

+(

0.025/√

3

2

)2

= 0.007

where VP2 is the volume taken from the secondarystandard solution for preparing each calibration point


and VF3 the final volume of such solutions.

u(VP2) = correction√3

+ CV nominal

= −0.44√3

+ 0.0026× 100= 0.006

In this case the correction and coefficient variationare also the mean value of the interval to which theybelong.

Finally, the uncertainty associated to the preparationof the calibration standard solutions is

urel(std) =√

u2rel(prim–secon) + u2

rel(dil)

=√

(0.039)2 + (0.007)2 = 0.039

u(std) = 0.039× 0.5 mg/l = 0.020 mg/l

4.5.2.1.2. Estimation of u2(cal). The uncertaintydue to the transformation of chromatographic signalsin concentrations is estimated by applying the expres-sion for the linear regression of least squares of resid-uals [15] (Table 2)

u(cal)

= 1

b

√s2resid

1

n+(Ci − C)2s2

b

= 1

7.250

√(0.146)2 1

15+ (0.500− 0.583)2(0.121)2

= 0.005 mg/l

urel(Fdil) =√

u2rel(VP3 + VP4) + u2

rel(VP3)

=

√√√√√√√

(0.007/√

3)2 + (0.015/√

3)2

1 + 3

2

+(

0.007/√

3

3

)2

= 0.005

whereb is the slope of the calibration curve,sb its stan-dard deviation,sresid the standard deviation of resid-uals,Ci the analyte concentration at each calibrationlevel, andC the average concentration.

4.5.2.1.3. Estimation of u2(repet). In order to es-timate the uncertainty associated to the precision, 10

aliquots of a sample spiked at 0.5 mg/l were analysedin repeatability conditions. This uncertainty is givenby the expression

u(repet) = ss√r

= 0.019√1

= 0.019 mg/l

wheress is the standard deviation from the chromato-graphic signals andr the number of replicates of eachsample when analysed in routine analysis (Table 2).

Thus the uncertainty derived from the estimation ofthe analyte concentration from the calibration curve,u(CA), is given by

u(CA) =√

u2(std) + u2(cal) + u2(repet)

=√

(0.020)2 + (0.005)2 + (0.019)2

u(CA) = 0.028 mg/l,

urel(CA) = 0.028 mg/l

0.5 mg/l= 0.056

4.5.2.2. Estimation of the uncertainty derived from thedilution of the sample extract u(Fdil). This uncer-tainty component is present only in the case of ECDanalysis because, NPD analyses are performed with-out dilution of the sample extract (Fig. 1B).

The dilution factor is calculated as

Fdil = Vfinal

Vinic= VP3 + VP4

VP3= 1 + 3

1= 4

where VP3 is the sample extract volume taken fordiluting and VP4 then-hexane volume added (Table 4).The associated uncertainty is

4.5.2.3. Estimation of the uncertainty derived fromthe sample concentration, u(CS).The sample con-centration in the final extract is given by the quotientbetween the sample weightmS and the volume of ex-tractVS

CS= mS

VS= 50 g

5 ml= 10 g/ml = 10 kg/l


and its uncertainty is calculated as (Table 4)

urel(CS) =√

u2rel(mS) + u2

rel(VS)

=

√√√√(0.033/√

3

50

)2

+(

0.015/√

3

5

)2

urel(CS) = 0.002

4.5.3. Combined uncertaintyOnce calculated, the relative standard uncertainty of

each uncertainty source, the overall combined uncer-tainty of the analytical method can be estimated fromthe general expression stated above.

urel(CON) =√

u2rel(CA) + u2

rel(Fdil) + u2rel(CS)

=√

(0.056)2 + (0.005)2 + (0.002)2

= 0.056

4.5.4. The correction of resultsUsually, multiresidue analytical methods yield re-

covery factors different than 100% for a wide rangeof analytes. This led to two alternatives: to perform acorrection of the obtained concentration value on thebasis of the recovery of the analyte or to incorporatethe recovery as another contribution to the combineduncertainty. To discuss the appropriate decision it isnecessary to test the signification of the bias in func-tion of the obtained recovery factors.

4.5.4.1. Testing the trueness of the analytical method.Bias is the difference between the estimated concen-tration CA and the actual concentration while recov-ery (R) (Table 2) is the quotient between the estimatedconcentration and the actual concentration, thus biascan be expressed as

Bias =(

1 − 1

R

)CA =

(1 − 1

0.818

)0.409

= −0.091 mg/l

and the uncertainty associated to bias is obtained with

the expression

u(Bias)

=√

CA2

R4u2(R) +

(1 − 1

R

)2

u2(CA)

=√

(0.409)2

(0.818)4(0.015)2 +

(1 − 1

0.818

)2

(0.028)2

= 0.011 mg/l

whereu(R) is the uncertainty associated to recovery,which depends on the tolerance of the reference ma-terial u(tol–stand) and on the standard deviation of re-covery factors divided by the square root of the num-ber of replicates,u(prec–inter):

u(R) =√

u2(tol–stand) + u2(prec–inter)

=√(

0.001√3

)2

+(

0.048√10

)2

u(R) = 0.015

In order to test the trueness, a test is used for verifyingif the ratio between the value of the bias and its uncer-tainty, is greater (significant) than the correspondingcoverage factork. Depending on the result of the test,two situations can be noted:

1. Bias not significant “NS”: the correction of resultsis not necessary, the termu(R) can be included inthe equation of the combined uncertainty:

CONNS = CON ⇔ urel(CON)NS

=√

u2rel(CON) + u2

rel(R)

2. Bias significant: in this case, it is necessary to de-cide whether correcting or not, the analyte contentin the samples due to the recovery:2.1. In the case that correcting results is chosen

“C”, the R.S.D. is included in the global equa-tion:

CONC = CON

R⇔ urel(CON)C

=√

u2rel(CON) + u2

rel(R)


CONC = 0.409

0.818= 0.500 mg/kg

⇔ urel(CON)C = 0.058

2.2. If the decision is not correcting “NC”, the rel-ative value of the bias is considered as a con-tribution to the combined uncertainty.

CONNC = CON ⇔ urel(CON)NC

=√

u2rel(CON) + (Bias)2

rel

CONNC = 0.409 mg/kg ⇔ urel(CON)NC

= 0.229

Therefore in�-HCH, the test results that theratio is greater than the coverage factor (k =2) and the estimated concentration is statisti-cally different to the actual concentration.

|Bias|u(Bias)

= | − 0.091|0.011

= 8.1 > 2

⇒ Bias significant

4.5.5. Estimation of the combined expandeduncertainty

The expanded uncertaintyU, is calculated as theproduct between the combined uncertaintyurel(CON)(bias not significant “NS”, bias significant and cor-rected “C” and bias significant and not corrected“NC”) and the coverage factork,

Urel(CON) = k urel(CON)

k = 2 considering that the uncertainty estimation isobtained from at least 10 replicates of the measurand(νef ≥ 10), which provides a coverage probability of95%. If the number of replicates is less than 10, itshould be used as an estimated degrees of freedomobtained from the equation of Welch–Satterwhaite:

νef = u4∑iu

4i (x)/νi

whereu is the combined uncertainty,ui(x) each oneof the significant components of the uncertainty andνi the degrees of freedom.

In the case of�-HCH, where the decision is correct,the result is calculated as (k = 2):

CONC ± U(CON)C = 0.500± 0.232 mg/kg

Urel(CON)C = 11.6%

Fig. 5 shows that the most significant component ofthe uncertainty is due to the estimation of the�-HCHconcentration from the calibration curve (u(CA)). Itis also observed that this component increases as theconcentration of standards decreases, due to the vol-ume taken from the secondary standard solution forpreparing the calibration solutions also decreases. Onthe other hand, since the bias is significant, it canbe noted that when the pesticide content is corrected,the associate uncertainty (u(CON)C) is minor than theassociate uncertainty (u(CON)NC) when the pesticidecontent is not corrected.

4.5.6. Factors influencing the uncertaintyFigs. 6 and 7 show the uncertainty calculated for

each identified source,u(CA), u(Fdil ) and u(CS). Itcan be observed that the most influencing factor inthe combined uncertainty is this associated to thepreparation of standard solutionsu(CA). Consideringu(CA), the contribution of the uncertainty associatedto the calibration curve is not significant comparedwith the contribution of the uncertainties associatedto the preparation of the primary standard solutionsand to the precision. As an example the combineduncertainty for achrinathrin is about 40 times greaterthan the obtained for most of pesticides (Table 5),mainly due to the small quantity of solid standardmeasured for preparing the primary calibration solu-tion (3.0 mg, Table 1) and the high standard deviationof recovery factors (Table 2).

Theu(prec) is affected by the spiking level in whichthe precision has been obtained, which is usually es-tablished on the basis of the MRL regulated for eachpesticide in each commodity (Table 2). It can be seenin the case of chlozolinate, nuarimol and triadimefonthat this contribution is the most important to the com-bined uncertainty. For rest of the analytes the contri-bution of the standard preparation and of the precisionto the combined uncertainty is similar as can be seenin Figs. 6 and 7.

In general NPD analysis shows a greater uncertaintylevel than ECD analysis, mainly due to the MRLs


Fig. 5. Diagram of the different uncertainty components for each standard concentration for�-HCH.

Fig. 6. Diagram of relative uncertainty of the pesticide analysed by GC–ECD.


Fig. 7. Diagram of relative uncertainty of the pesticide analysed by GC–NPD.

established for OP pesticides are usually minor thanthe established for OC pesticides and also due to theprecision of NPD is usually minor than the precisionof ECD measurements.

Other contributions such asu(Fdil ) and u(CS) areless significant compared with the above factors.

4.5.7. The correction from recovery factorsTable 5 shows the combined uncertainty for each

analyte at the three concentration levels used for thecalibration curves, including the obtained when thecorrection from recovery factors is carried out andwhen the correction from recovery factors is not per-formed. Previously, it has been stated whether the biasis significant or not with at-test, which compares therecovery factors obtained with 100%. When recover-ies have not a significant bias, the uncertainty consid-ered is the same than when correction is carried out. Itcan be seen that the uncertainty is strongly dependingupon the concentration level, being the greater thesecalculated at the first concentration level, which is usu-

ally established on the basis of MRLs requirements.This is the case of acrinathrin, triadimefon, cipro-conazole, buprofezin, methamidofos and myclobu-tanil among others, which show uncertainties higherthan 100%.

Considering the correction of the bias when it issignificant, it can be observed an increasing of the un-certainty, due to the inclusion of the relative value ofthe bias as a contribution to the combined uncertainty.In GC–ECD analysis, the uncertainty with correction,at the lowest concentration level, range between 12.7and 25.9% (excluding uncertainties higher than 100%,Table 5); these levels increase to 30.9–64.2% if thecorrection is not performed. It can also be observedthat this effect is greater when the systematic error isgreater, low recovery factors with low R.S.D., as anexample dichlofluanid shows a combined uncertaintyminor than 12.7% with correction, and increase toaround 64% when the correction is not considered (re-covery factors of dichlofluanid was 76.1% with 1.7%R.S.D., Table 2).


Table 5Combined uncertainty at the different concentration levelsa

Pesticide Bias not significanturel(CON)NS (%) Bias significant

Level 1 Level 2 Level 3 Correctedurel(CON)C (%) Not correctedurel(CON)NC (%)

Level 1 Level 2 Level 3 Level 1 Level 2 Level 3

ECD�-HCH 18.7 11.8 9.9 48.1 45.9 45.4Chlorothalonil 23.3 15.0 12.2 64.2 61.7 61.1Vinclozolin 18.5 13.2 11.5 44.7 42.8 42.3Dichlofluanid 12.7 10.8 10.4 64.0 63.6 63.6Triadimefon 168.6 84.9 43.7Chlozolinate 100.7 51.2 27.2 109.2 63.3 50.3Procymidone 15.2 10.5 9.5 30.9 28.9 28.6Endosulfan-� 18.2 13.0 11.4 30.9 28.1 27.4Endosulfan-� 25.1 14.5 10.5 40.3 34.8 33.3Endosulfan-s 16.5 11.6 10.1 32.6 30.3 29.8Nuarimol 128.6 65.0 33.9Iprodione 25.0 15.5 12.5 36.9 31.2 29.8Bromopropylate 19.7 13.4 11.5 29.5 25.7 24.8Tetradifon 25.9 16.3 13.0 31.3 26.3 24.4Acrinathrin 985.0 493.0 105.0

NPDDichlorvos 36.4 13.5 10.4 43.0 26.5 25.1Methamidofos 480.7 160.8 97.4Acephate 99.2 35.1 23.5 102.3 43.1 34.3Heptenophos 74.0 27.7 19.9 78.7 38.6 33.5Ethoprophos 436.2 145.9 88.2Dimetoate 37.5 15.7 13.1 44.8 29.0 27.7Diazinon 224.7 76.1 47.1Parathion-methyl 51.7 19.8 15.1 60.3 36.8 34.5Chlorpyrifos-methyl 42.5 17.3 14.2 55.3 39.3 38.1Pirimiphos-methyl 41.7 20.8 18.9Malathion 50.5 20.0 16.6 56.8 32.7 30.7Parathion-ethyl 42.5 16.7 14.3 55.0 38.6 37.6Chlorpyrifos 56.7 22.8 18.5Chlorfenvinphos 41.2 18.0 15.6 51.2 35.3 34.2Triadimenol 60.8 27.2 23.1Fenamiphos 35.3 19.0 18.1Myclobutanil 360.9 121.9 75.1Buprofezin 398.9 67.5 42.1Cyproconazole 384.4 128.7 78.0Ethion 45.9 18.6 16.1 52.2 30.9 29.5Carbophenothion 312.5 105.5 64.1

a Levels 1–3 are the concentration levels of the calibration curve given in Table 4.

4.5.8. Strategies for decrease the uncertaintyFigs. 6 and 7 show that pesticides as chlozoli-

nate, nuarimol, triadimefon, diazinon, cyproconazole,carbofenothion, buprofezin, dimetoate, ethoprophos,methamidofos and myclobutanil present a componentof the uncertainty due to the precision,u(prec), veryimportant, this fact makes that the global uncertainty

has a high value with regard to the rest of the pes-ticides. In addition most of these pesticides requireconcentration levels for the calibration curves verylow, and therefore when the relative uncertainties arecompared, the value of these increases, although thecomponent of the uncertainty due to the calibrationcurves is not significant.


On the other hand, the uncertainty associated to thepreparation of the standards,u(std), is quite similar inall pesticides and higher than other components of theglobal uncertainty because the amount of solid stan-dard weighted for primary standard solution solutionswas very little.

Therefore in order to decrease the uncertainty ofthe analytical method it would be convenient to actson the two components previously mentioned, on onehand trying to diminish the precision of the method orto increase the concentration levels and on the otherhand increasing the amount of solid standard weightedfor the preparation of primary standard solutions.

5. Conclusions

A methodology for calculating the uncertainty ofresults on the basis of in-house validation data hasbeen applied to a pesticide multiresidue method. Un-certainty sources have been identified and standarduncertainty established.

The most significant uncertainty sources are thepreparation of the standard solutions, mainly theweigh-out step for preparing the primary standard so-lution, and the precision of the method. In this sense,when less than 10 mg of solid standard is measuredfor preparing the primary standard solution, the un-certainty increases dramatically to values higher than100%.

ECD determinations show in general less un-certainty than NPD analysis due to the MRL fororganochlorine pesticides are usually greater than fororganophosphate pesticides, so that the concentrationlevel is another factor which influences very muchthe uncertainty level. This fact is because the un-certainty associated to the precision increase at lowconcentration levels.

When a systematic error is present, the bias is signif-icant, a correction of recovery factors would decreasethe uncertainty of results dramatically.

MRLs established at very low levels, close to0.05 mg/kg, are usually determined with uncertaintiesaround 300%, which means that the analytical methodis semi-quantitative at this concentration level. So it

would be difficult to establish whether a sample ispositive or negative when such pesticides are detectedunless an uncertainty level be established by regulatorynorms.

Acknowledgements

The authors are grateful to Institute National ofInvestigation and Agrarian and Alimentary Technol-ogy (INIA), Ministerio de Agricultura, Pesca y Al-imentación(Project CAL00-002-C2-2) for financialsupport.

References

[1] M. Valcárcel, M.D. Luque de Castro, A hierarchical approachto analytical chemistry, Trends Anal. Chem. 14 (1995) 242.

[2] A.M. Garcıa-Campaña, J.M. Bosque Sendra, L. Cuadros-Rodriguez, E. Almansa López, Biomed. Chromatogr. 14(2000) 27.

[3] ISO, International Vocabulary of Basic and General StandardTerms in Metrology, International Standards Organisation,Geneva, 1993.

[4] ISO, Guide to the Expression of Uncertainty in Measurement,International Standards Organisation, Geneva, 1993.

[5] EURACHEM Guide, Quantifying Uncertainty in AnalyticalMeasurement, 2nd Edition,http://www.vtt.fi/ket/eurachem/quam2000-p1.pdf.

[6] Analytical Methods Committee, Analyst 120 (1995) 2303.[7] Protocol for In-house Method Validation, International Union

of Pure and Applied Chemistry (IUPAC), 1999.[8] A.R. Hill, S.L. Reynolds, Analyst 124 (1999) 953.[9] A. Maroto, R. Boqué, J. Riu, F.X. Rius, Trends Anal. Chem.

18 (1999) 577.[10] A. Maroto, R. Boqué, J. Riu, F.X. Rius, Quım. Anal. 19

(2000) 85.[11] R.J.N. Bettencourt da Silva, M. Joäo Lino, J.R. Santos,

M.F.G.F.C. Camöes, Analyst 125 (2000) 1459.[12] Harmonised Guidelines for the use of recovery Information

in Analytical Measurements, IUPAC Technical Report, PureAppl. Chem. 71 (1999) 337.

[13] Quality Control procedures for pesticide residue analysis,Guidelines for residues monitoring in the European Union,Document 7826/VI/97, European Commission, Brussels,1997.

[14] F.J. Egea González, J.L. Martınez Vidal, M.L. Castro Cano,M. Martınez Galera, J. Chromatogr. A 829 (1998) 251.

[15] L. Cuadros Rodrıguez, A.M. Garcıa Campaña, J.M. BosqueSendra, Anal. Lett. 29 (1996) 1231.

http://www.vtt.fi/ket/eurachem/ quam2000-p1.pdf

Assessment of uncertainty in pesticide multiresidue ...hera.ugr.es/doi/15085983.pdf · carrier gas used was helium (purity 99.999%). Mass spectrometer settings: solvent delay, 4.5min;

Documents