tropical crop consultants limited · 2018. 11. 23. · 1. use of leaf sampling unit (lsu) blocks in...

TropicalCropConsultantsLimited

Minimizationoferrorinleafanalysissamplingsandanalysis1 ThomasFairhurst2

Researchhasshownthat,whencarriedoutcorrectly,leafanalysisprovidesagoodindicationofpalmnutritional status. All plantations therefore place great emphasis on the results of leaf and rachissampling and analysis as the basis for determining fertilizer recommendations to correct nutrientdeficiencies. Because fertilizers represent such a large proportion of operational costs and have amajor impact on yield, it is important that leaf sampling and analysis is based on correctlyimplemented procedures. Yet many plantations take insufficient care to verify that laboratoryanalysisisaccurateandprecise.

Leafsamplingandanalysisprocessinvolvessixmainsteps,eachofwhichispronetoerror.Errorsinthe leaf sampling and analysis process (Elsa), are cumulative andmay arise from a combination oferrors in LSUblock setup (Eblock), LSUpalm identification (Epalm), selectionof leaf 17 (Eleaf), samplepreparationattheplantation(Eprep),laboratoryerror(Eanal),anddataprocessing(Edata).

Itisessentialtominimizetheerrorsateachstageoftheprocessinordertoachieverepresentative,accurate and precise results that can then be used to determine fertilizer requirements.Withoutproper procedures and cross checks, errormaybe as great as 10–20%, leading to ‘false positives’(i.e., deficiencies reported where none exist) and ‘false negatives’ (deficiencies exist but are notreported).Bothwillaffecttheamountoffertilizerrecommended,yieldsandprofits.

Inthispaperwereviewcommonsourcesoferrorinthesixstagesintheprocessofleafsamplingandanalysisandhowtheycanbecontrolled.

1. Useofleafsamplingunit(LSU)blocksIn the past, leaf sampling units (LSUs)were defined as ‘a set of blocks so similar in terms of soil,planting material, palm age and field conditions that it can be considered a single unit for thepurposeoffertilizerrecommendations’.LeafsamplingwasthencarriedoutinasingleblockthatwasconsideredrepresentativefortherespectiveLSU.

Nowadays itmakesmore sense to assess palmnutrient status and fertilizer requirements in eachblockbecause:

� Thecostofleafsamplingandanalysisissmallbycomparisonwiththecostoffertilizers.

� Block-by-blockassessmentprovidesthemeansforgreaterdiscriminationonfertilizerrates.

2. IdentificationofLSUpalmsTheresultsofleafanalysismaybedistortedifthereareerrorsinpalmsamplingprocedures:

Ifdifferentpalmsaresampledeachyear, thedifferences in leafnutrientstatusbetweenyearsandblocks may be due to palm selection rather than block nutritional status. For this reason, leafsamplingshouldbecarriedoutonagridofpermanentlyselectedandmarkedLSUpalms(Table1).

1ExtensionBulletinTCCL-001,4November20152 TropicalCropConsultantsLimited,26OxenturnRoad,Wye,Kent,TN255BE,UnitedKingdom·[email protected]·www.tropcropconsult.com·+441233811873

Elsa = Eblock +Epalm +Eleaf +Eprep +Eanal +Edata

Minimizationoferrorinleafanalysissamplingsandanalysis

Page2

� LSU palms aremarked in triangular pattern, every nth palm in every nth row (Figure 1). Thetriangular pattern providesmore evendistribution of LSUpalms andmakes sampling easierandmoreefficient.

Table1. Leafsamplingunitpalmselectionbyplantingsystem.Plantingsystem SamplinggridFlat lined planting where most blocks arerectangularshaped

Triangular grid with every 10th palm in everytenthrow.

Contourplanting Every 10th palm along every 10th contourterrace.

Low-lyingareaswithdrainseveryfourpalmrows Every8thpalminevery8throw.

Figure1. LocationofLSUpalmsinarectangular-shapedblockplantedonflatland.

� LSUpalmsmustbeatleastfourpalmsdistantfromtheroadside(toavoid‘roadsidebias’andcontaminationfromdust).

� Only healthy and productive palms should be selected as LSU palms (because we are notinterested in thenutritional statusof abnormal andunproductivepalms).Make sure sterile,abnormal,leaning,supply,anddiseasedpalmsarenotselectedasLSUpalms.

� AllLSUpalmsaremarked,georeferencedandmappedonisometricpaperand/ordigitalmaps.LaminatedmapscanbeusedinthefieldtoassistwithLSUpalmidentification.

� AllLSUpalmsarelabelledwithaQRcode.Thesamplingteam‘read’theQRcardwithahandhelddevicetoverifythatthepalmhasbeenvisitedduringleafsampling.

� RecordthenumberofLSUpalmsineachblockandverifythatthesamplingdensityisateast1%:

5 15 25

Collection road Main road Palm occupied by road Palm LSU palmPath taken by leaf sampling team 5 Leaf sampling row


Page3

3. SelectionofreferenceleaftissueLeaves on theoil palmare arranged in eight parastichies and eachparastichy contains 5–8 leaves(Figure 2). The first fully open leaf is Leaf 1 and the 6th leaf in the 1st parastichy is Leaf 41. Thereferenceleafforleafanalysishasbeennominatedbasedonathoroughanalysisofthephysiologicalcharacteristicsoftheoilpalmleafcanopy:

� Leaf9isusedinimmaturepalms<3YAP

� Leaf17(locatedinthemiddleoftheleafcanopy)isusedinmaturepalms≥3YAP.

Figure2. Phyllotaxisinleft-handed(left)andright-handed(right)palms.Leavesinonespiral(e.g.,Leaf1,9,17,25,33,and41)constituteoneparastichy.

Because LSUpalms aremarked, it is easy for a supervisor to check that the correct leaf hasbeensampled by spot-checking a few LSU palms in several blocks immediately after leaf sampling hasbeencompleted.

Maturepalmsproduceabout1.5–2.5leavespermonthor18–30leavesperyearandthereforeLeaf17is7–11monthsoldandLeaf9is4–6monthsold.

Thereferenceleafpinnaetissueonthereferenceleafis’12leafletsperleaf(i.e.,6leaflets(3upperand3 lower rank) fromeach sideofapointon the rachis⅔rdsof thedistance form the insertionpointofthefirsttrueleavesandtheleaftip’.Thecorrectpositioncoincideswiththepointwheretheridgeontheuppersurfaceoftherachistaperstoapoint(Photo1).

Incorrect identification of reference leaf tissue leads to ‘false positives’ and ‘false negatives’concerning leafnutrientdeficiencies.Forexample, leafN,PandKcontentdecreaseswhilst leafCacontentincreaseswithleafage,whichhasimportantconsequences:

� If the leafselected isyounger thanthenominatedreference leaf, the leafcontentofN,P,KandMgwillbegreater,leadingtoafalseconclusionthatpalmnutrientstatusissufficient.

� Iftheleafselectedisolderthanthenominatedreferenceleaf,theleafcontentofN,P,K,andMgwillbesmaller,leadingtoafalseconclusionthatpalmnutrientstatusisdeficient.

Leafnutrientcontentalsovarieswithinleafpinnae.Itisthereforeimportanttoselectthereferencetissueofeachleaf(i.e.,themiddlethirdportionofeachleaflet).Theleafpinnaemidribshouldalsoberemovedanddiscarded.

Accurate and rapid identification of the reference leaf (Leaf 9 or Leaf 17) requires practice. It isrecommendedthattheagronomistattachalabeltoallleavesinthepalmcanopyonayoungmaturepalm(<8YAP)inanaccessiblelocation.Numberedplastictags(cutoutofausedherbicidecontainer)

Samplingdensity=NumberofLSUpalms

Density(palm/ha)×area(ha)


Page4

areattachedtotherespectiveleafrachiswitharubberband(Photo2).Leafnumberchangesasthepalmgrowsandthereforethelabelsshouldbe‘moved’everytwomonths.

Photo1. Leafletsaresampledfromeachsideofapointontherachis⅔rdsofthedistanceformtheinsertionpointofthefirsttrueleavesandtheleaftipareselectedforleafanalysis.

Photo2. AllleafpositionshavebeenlabeledonayoungpalmandtherelativepositionofLeaf1,9,17,25,33and41isrevealed.

1 2 3 4 5 6

1 2 3 4 5 6


Page5

This exercise helps to ensure that all personnel know how to identify Leaf 9 and Leaf 17 (thereference leaves for leaf sampling) and Leaf41 (the leaf subtending thenext ripebunch,used formeasuring palm height). Labeling all leaves in the palm canopy also helps agronomy andmanagement staff to understand pruning standards (i.e., number of leaves and leaf parastichiesrequiredinrelationtopalmage)andhowtoidentifythepresentpositionofLeaf1identifiedatthepreviousleafsamplingexercise.

Arefreshercourseonleafsamplingproceduresshouldbecarriedouteachyearbeforethestartofleafsampling.

4. SamplepreparationAllleafsamplesshouldbedriedandgroundattheplantation.Otherwiseitisimpossibletodisguisecrosschecksampleswithinabatchofcommercialsamples.

Somestepsinsamplepreparationaffectleafnutrientcontent:

� Samplesmust be dried <24 hours after sampling. LeafN content is affected if leaf drying isdelayed.

� Leafsamplesshouldbedriedinaforceddraftovenat70–80oC.LeafNcontentisreducedifleavesaredriedat>80oC.

� Driedsamplesaregroundtopassa18-mesh(or1mm)sieveusinge.g.,aThomasWileymilloraTecatorCyclotecgrinder.Uniformparticlesizeisimportantbecauseaverysmallamountofsample(±0.5g)fromtheblocksample(600g)isusedforanalysis(Figure3).

Averysmallamountofsample(0.2–0.5g) isusedforeachanalysis.Therefore,makesurethatthesamplesenttothelaboratory(±50g)isrepresentativeoftheblocksample(±600g)(Figure3).

Topreparearepresentativesub-sample:

� Thoroughlymixtheblocksampleandplaceitonacleanplasticsheet.

� Formthesampleintoaconeanddividetheconeintofourquadrants.

� Usea5ml laboratory spoonand takeonespoonful fromeachquadrantof thecone in turnuntiltheplasticsamplebagcontains50gofsamplematerial.

Figure3. Relativesizeofablockleafsample(600g),samplesenttothelaboratory(50g)andsampleusedforanutrientdetermination(0.5g)

Samplesarepackedinlabelledplasticbagsbeforedispatchtothelab.

5. AnalysisataservicelaboratoryLaboratorycross checking isessential,even if the laboratory isownedby theplantationcompany,whenevertheresultsofleafanalysisareusedtodeterminepalmnutrientstatusforthepurposeofmakingfertilizerrecommendations.

Thekeyissuesforaccurateandpreciselaboratoryworkare:

� Well-maintainedequipment.

� Highqualityreagents.

Block leaf sample Sent to laboratory for analysis Used to determine nutrient content


Page6

� Skilled,experiencedanddiligentanalysts.

These aspectsof laboratorymanagement are, of course, outside the supervisionof theplantationcompany.However,anobservantrepresentativeoftheplantationcangetafairlyclearimpressionofa service laboratory during a visit. As a general rule, a busy laboratorywill provide better resultsbecausetheanalystsarepractisingtheirskillseachday.

5.1. WEPALparticipation

TheselectedservicelaboratoryshouldbeaparticipantintheWageningenEvaluatingProgrammeforAnalytical Laboratories (WEPAL), run by the University of Wageningen. WEPAL requires memberlaboratories to analyze blind sampleswith acceptable levels of accuracy and precision in order tomaintain accreditation. However, WEPAL only verifies whether or not a laboratory is capable ofaccurate and precise analysis – it does not necessarily mean that the laboratory will performaccurateandpreciseanalysisonasetofcommercialsamples!

5.2. Crosscheckingforerrors

Errorsinlaboratoryanalysiscanleadtosignificantandcostlymistakesinfertilizerrecommendations.Therefore,whilsta service laboratorymay run theirown internalquality control cross checks,andparticipate in the WEPAL certification scheme, the plantation should run independent ‘external’crosschecksontheservicelabtoverify,witheachbatchofcommercialsamples,thatanalyticalworkmeetstherequiredstandards.

Itisimportanttoassessbothaccuracyandprecisionoflaboratoryanalysis:

� Analysisisaccuratewhenthevaluereporteddoesnotdifferfromthe‘truevalue’.Accuracyisthenearnessofameasurementorthemeanofasetofmeasurementstothetruevalue(i.e.,theknownacceptedvalueofaquantifiableproperty).

� Analysisisprecisewhenmultipleanalysesofagivensampleagreewitheachother.Precisionissometimes referred to as ‘reproducibility’ and is ameasure of the random error associatedwithaseriesofrepeatedmeasurementsofthesameparameterwithinasample.

Precisionandaccuracyarespecifictotheanalystandinstrumentandchecksarethereforerequiredwhentheanalystischangedornewinstrumentsareinstalled.

ExamplesofaccuracyandprecisionareillustratedinFigure4.

Figure4. Examplesofprecisionandaccuracyintheanalysisofsevenidenticalsamples.

Resultsmaybe:

4) Not precise and not accurate

2) Accurate but not precise1) Accurate and precise

3) Precise but not accurate


Page7

� ‘Accurateandprecise’ (i.e., smallvariationbetweenseven identical samples, resultsclose to‘truevalue’,Figure1.1).

� ‘Accuratebutnotprecise’(i.e., largevariationbetweensevenidenticalsamples,resultscloseto‘truevalue’(Figure1.2).

� ‘Precisebutnotaccurate’(i.e.,smallvariationbetweenidenticalsamplesbutresultsdifferentfrom‘truevalue’,Figure1.3).

� ‘Not precise and not accurate’ (i.e., large variation between identical samples and resultsdifferentfrom‘truevalue’,Figure1.4).

Wewillnowdetailmethodstoassesslaboratoryaccuracyandprecision.

5.2.1. AssessmentofaccuracyTo assess accuracy, the plantation must prepare a standard leaf sample (SLS) with known ‘truevalues’fortheconcentrationofallnutrients(N,P,K,Mg,Ca,B,Cl,Mn,Zn).TheSLScanbepreparedbysamplingLeaf17on1,000palmsintheusualway.Samplesarethendried,ground(1mmor18mesh sieve) and thoroughly mixed to produce a uniform and homogenous sample of groundmaterial.About5kgofSLS shouldbepreparedsince the standard sample shouldbe sufficient foraccuracycrosscheckingexercisescarriedoutover5–10years([email protected]=0.35kg/year).Thestandardsampleisstoredinanairtightglasscontainer.The‘truevalue’ofthissamplecanbedeterminedbyanalysisatahighlyreputedlaboratory.Sevenidenticalsamplesarepreparedfromthestandardsampleandsentforanalysis(N,P,K,Mg,B,Cu,Zn,Mn)atareliablebenchmarklaboratory(e.g.,UCDavis,WEPAL,Rothamsted).Themeanvalueforeachnutrientcanbeusedasthe‘truevalue’forthestandardsamplewhenassessingtheaccuracyofservicelaboratories.

Theminimumdetectionlevel(MDL)providesanindicationoftheminimumamountofanalytethatthelabcandetect.MDLiscalculatedasfollows:

WhereSisthestandarddeviationforthesevensamples.Thetvaluefor99%and6degreesoffreedom(n=7,7-1=6)is3.143.

In a hypothetical example, seven samples were sent to a research laboratory to determine ‘truevalue’fortheNcontentoftheSLS.Themeanvaluewas2.50±0.016%Nor2.49–2.52%NandtheMDLwas0.05(Table2).

An identicalsetofsevensampleswassenttoacandidateservice laboratorytocheckonaccuracy.Themeanvaluewas2.62±0.057%Nor2.57–2.68%NandtheMDLwas0.19(Table2).Basedonthisassessment the candidate service laboratory was rejected because the analysis was insufficientlyaccurate.

Results can be plotted in a graph showing themean value and confidence interval for the sevenstandardleafsamples(SLS)andthe‘truevalue’forthestandardsample(FigureXa).Inthisexample,service laboratoryperformancewasgood in2012. Themean from the sevenSLSwas close to the‘true value’, indicating good accuracy and the calculated confidence interval was small, indicatinggood precision. In 2013 the mean from the SLS was also close to the true value but variabilityamongst the seven SLSwashigh, indicating acceptable accuracybut lackof precision, In 2014 themean value for the SLS was greater than the ‘true value’, indicating lack of accuracy but theconfidenceintervalsweresmallindicatingacceptableprecision.In2015bothaccuracyandprecisionwaspoor.

MDL= t(6,0.01%) × S


Page8

Table2. Hypothetical example of analysis of standard leaf material and the calculation of theminimumdetectionlevel.Sample # Formula LeafN%

Researchlab

Servicelab

1 2.50 2.632 2.52 2.523 2.48 2.684 2.49 2.705 2.51 2.586 2.50 2.617 2.53 2.65Numberofsamples(n) a =COUNT(S1:S9) 7.00 7.00Mean(x) b =AVERAGE(S1:S9) 2.50 2.62Standarddeviation(S) c =STDEV.S(S1:S9) 0.0172 0.0613Samplevariance(S2) d c2 0.0003 0.0038Standarderror(SE) e =√d/a-1 0.006 0.023t-value(t(6,2.5%)) f t(6,2.5%) 2.447 2.447Confidenceinterval(CI(95%)) g exf 0.016 0.057Upperlimit h b+g 2.52 2.68Lowerlimit i b-g 2.49 2.57Minimumdetectionlimit(MDL) j 3.143xc 0.05 0.19Maximumrelativepercentdifference(RPDmax) k 2 7

5.2.2. Assessmentofprecision

Precision can be assessed by measuring the difference between the values obtained for pairs ofidenticalsamples.Forthispurpose,fivepairedsamplesaresentforanalysis.Thepairedsamplescanbepreparedbydividingfivecommercialsamplesintotwoidenticalsub-samples.Thefivecommercialsamples selected for this exercise should represent a rage of nutrient status (i.e., from deficientblocks to blocks with good nutritional status, based on past analysis) so that precision can beassessedatsmallandlargenutrientconcentrations.

Anestimateofprecision is providedby the relativepercentdifference (RPD). The relativepercentdifference(RPD)iscalculatedasfollows:

Whereaandbareasamplepair.

Forexample,ifsamplea=2.50%Nandsampleb=2.45%N,

LowervaluesforRPDindicategreaterprecisionand,formostanalysis,theRPDshouldbe<5%.ReallygoodlaboratoriesachieveRPDof<2%.

Relativepercentdifference= a−ba+b( ) /2

⎛

⎝⎜⎜

⎞

⎠⎟⎟x100

RPD= 2.50−2.452.50+2.45( ) /2

⎛

⎝⎜⎜

⎞

⎠⎟⎟x100

= 0.052.48

x100=2%


Page9

Thestandarddeviationofdifferencesbetweenpairedsamplesprovidesanindicationofprecisioninthebatchoffivepairedsamples:

Where S = standarddeviationof thedifferencebetweenduplicatepairs,N=numberof duplicatepairsand(x1-x2)i=differenceofthe ithpair.Thegreaterthevalueofthestandarddeviationofthedifferencebetweenduplicatepairs,thepoorertheprecisioninanalysis.

In a hypothetical example, we can compare the RPD for five paired samples analyzed in twolaboratories (Table3). In thisexample, theRPDvalues for Lab1 ranged from1–2, indicatinggoodprecisionand the standarddeviationofdifferenceswas0.20.Bycontrast, in Lab2 theRPDvaluesrangefrom3–16,indicatingpoorprecisionandthestandarddeviationofdifferenceswas0.204

Table3. Resultsfortheanalysisoffivepairedsamplesintwoservicelaboratories.

Sample Lab1 Lab2

A B RPD (a-b)2 A B RPD (a-b)2

S1 2.45 2.49 2 0.0016 2.45 2.53 3 0.0064

S2 2.35 2.40 2 0.0025 2.50 2.30 8 0.0400

S3 2.52 2.54 1 0.0004 2.70 3.00 11 0.0900

S4 2.86 2.79 2 0.0049 2.20 2.40 9 0.0400

S5 3.15 3.20 2 0.0025 3.30 2.80 16 0.2500

Average 2.67 2.68 2 0.0024 2.63 2.61 9 0.0853

Standarddeviation 0.332 0.323 0.2041 0.415 0.289 1.2216

Wecanconcludethatinourhypotheticalexample,Lab1providesahighdegreeofprecisionwhilsttheprecisionofLab2isinsufficient.

Resultsfromthetwolaboratoriescanbeplottedinagraphtoshowthedegreeofprecisionbetweenpairedsamples(Figure5b).TheresultsforLab1lineupcloselytoa1:1line,indicatinggoodprecisionatallconcentrationsofN.TheresultsforLab2arepoorlyalignedtoa1:1 line,particularlyathighconcentrationsofN,indicatingpoorprecision.

Figure5. GraphsshowingmeanNconcentrationforastandardleafsampleoverafour-yearperiod(Barsrepresent95%confidenceintervalsandthe‘truevalue’isshowninblue)(a)andprecisionbetweenfivepairsofsamplesanalyzedforleafNcontentattwolaboratories(b).

S=x1-x2( )

i

2

i=1

n

∑2n

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

�

��

��


Page10

5.3. Crosschecksamplelabelling

It is important that the laboratory is not providedwith any information on the provenance (e.g.,block ID, palm age, planting material, cross check sample) of samples. Instead, each sample islabelledwithasix-digit IDcode(e.g.,010123)anddetailsofsampleprovenancerestoredwiththesample IDcodebytheplantation inaspreadsheet. It isparticularly importanttotagallcrosschecksampleswithalabelsimilarinformatandappearancetocommercialsamples.

It is always best to include all cross check samples with routine commercial samples so that it isimpossible for the analyst to guess the provenance of the samples. The laboratory should beinformed,however, thateach setof commercial samples contains a setof cross check samples toassess laboratory accuracy and precision. This will encourage the laboratory to carry out theanalyticalworkdiligently.

5.4. Assessmentoflaboratory

Accuracyandprecision isgreater forsomeassays (e.g.,RPD±2%for leafNcontentmeasured inaC/N auto analyzer) than others (e.g., ±5% for leaf P content measured by colorimetric method).Therefore,assessmentsforaccuracyandprecisionshouldbecarriedoutforallnutrients(N,P,K,Mg,Ca,B,Cu,Zn,Mn).

Itmayalsobeworthwhile toprepareastandardrachissample (particularlysincethecontentofNandPismuchlowerandKmuchhigherthaninleafsamples)andastandardsoilsample.

Theassessmentofaccuracyandprecisionshouldbecarriedouteachyearandtheresultsmaintainedin a computer database. In this way, accuracy and precision can bemonitored conveniently overtime.

6. Dataprocessing6.1. Correctionofcommercialleafanalysisdataforseasonalvariation

Inmost locations, leafnutrientcontentvaries throughtheyear.Forexample, theconcentrationofnutrients in samples taken at the end of the dry season is often notably smaller than in samplestakenattheendofthewetseason.Datafromannual leafsamplingshouldthereforebecorrectedforseasonalfluctuations.CarryoutleafsamplinginX%ofblockseverytwomonths(TableX).

Table4. ResultsfrombimonthlysamplingforleafNSurveyblock LeafNFebruary 2.51April 2.41June 2.48August 2.57October 2.61December 2.53Average 2.52

Tocorrectdatafromcommercialblocks,addthedifferencebetweenthemeanvalueandthevaluefor themonth closest to theactual samplingmonth for the commercial block. For example, if thecommercial sample value for leaf N sampled in October was 2.48 %N, the corrected value iscalculatedasfollows.

Allleafandrachisdatashouldbecorrectedforseasonalvariation.

6.2. Transpositionalerrors

Transpositionalerrorsoftenarisewhendataistransferredfromlaboratoryreportstothecompanyagronomicdatabase.Toreducethechanceoferror:

CorrectedleafN%=2.48+(2.61-2.52)=2.48+0.09=2.57


Page11

� Usetwooperatorstoenterdata(i.e.,onetoreadoutthedataandtheothertoenterthedatainthecomputer).

� Implement double data entry (i.e., the same valuemust be entered twice) in the companyagronomicdatabase.

7. Conclusions� Atotalof17samples(5pairedsamplesand7standardleafsamples)willprovideahighdegree

ofreassuranceconcerningtheaccuracyandprecisionofaparticularservicelaboratory.

• Analternative laboratoryshouldbesoughtwhereaccuracyandprecisionarefoundtobepoor.

• Wherehighlevelsofaccuracyandprecisionareevident,theagronomistcanbeconfidentthat the results from leaf sampling and analysis can be interpreted rigorously for thepurposeofestimatingfertilizerrequirements.

� Acrosscheckingexerciseshouldbecarriedoutwithahighly reputedresearch laboratory todetermine ‘truevalues’ foreachanalyteandrepeatedeachyearwiththeservice laboratory,usingresultsfromtheresearchlaboratoryasabenchmark.

� The WEPAL cross checking system only verifies whether or not a laboratory is capable ofcarryingoutaccurateanalyticalwork.A laboratorycertifiedbyWEPALmightnotnecessarilydeliveraccurateandpreciseresultsforsamplespresentedbyacommercialcompany.

� Alwayscarryoutcrosschecksforaccuracyandprecisionbeforeselectingaservicelaboratory.

� Carry out cross checks for accuracy and precision each year to verify that the selectedlaboratoryhasmaintainedstandards.Bothaccuracyandprecisionmaychangewithchangesinequipment,personnel

� Prepareastandardsampleof leaf (andrachis)material.Sendseven identical samplesof thesampletoahighlyreputedlaboratorytodeterminethe‘truevalue’fornutrientcontent(N,P,K,Mg,Ca,B,Cu,Zn).

� Alwaysinformthelabthatcommercialsamplesincludeindistinguishablecrosschecksamples.Thiswillencouragethelaboratorytotakemorecarewiththeanalyticalwork.

� Keeparecordofallassessmentsofaccuracyandprecisionforthepurposeofcomparisonandtotrackaccuracyandprecisionovertime.

� Avoid thetemptationtoattributesignificanceto theresultsof leafanalysisgreater thantheaccuracyandprecisionindicatedbycrosscheckingexercises.

Table5. SummaryofthemaincausesoferrorinlaboratoryanalysisActivity SourceoferrorLSUblocks Whenblocksaregroupedinleafsamplingunits(LSUs)thesample

blockmaynotberepresentativeoftheotherblocksintheLSU.

LSUpalmselection UnproductiveorabnormalpalmsselectedasLSU.Samplingdensity(LSUpalms/ha)insufficient.LSUpalmsnotdistributedevenlywithinthesampleblock.

Leafandrachissampling Incorrectleafsampled.Samplesmislabeled.

Samplepreparation Oventemperaturetoohigh(>80oC).Leavescontaminatedbyunnecessaryand/orineffectivewashingbeforedrying.Samplesmislabeled.


Page12

Laboratoryanalysis InsufficientinternalqualityassuranceprogrammesIncorrectmethodused.Proceduresnotfollowedcorrectly.Faultyorpoorlymaintainedequipment.Poorcalibrationofequipmentandstandards.Poorqualityorcontaminatedreagents.

Dataprocessing Dataenteredincorrectly.Confusionoverlabellingleadstodataentryerrors.

Table6. AcronymsTerm DefinitionYAP YearafterplantingLSU Leaf samplingunit.A setofblocks so similar in termsofpalmage, soil type,planting

materialandagronomicconditionsthattheymaybeconsideredasasingleunitforthepurposeoffertilizerrecommendations.

YAP Yearsafterplanting.Theageofthepalm(calculatedasthepresentyearminusyearofplanting)

WEPAL WageningenEvaluatingProgrammeforAnalyticalLaboratoriesMDL MinimumdetectionlevelCI ConfidenceintervalSE StandarderrorS StandarddeviationS2 Samplevariancet Valuefromtableoft-valuesRPD Relativepercentdifference

tropical crop consultants limited · 2018. 11. 23. · 1. use of leaf sampling unit (lsu) blocks in...

Documents