Soilsbatj

Asoilparticle-sizedistributiondatasetforregionallandandclimatemodellinginChinaWeiShangguana,,YongjiuDaia,BaoyuanLiub,AizhongYea,HuaYuanaaStateKeyLaboratoryofLandSurfaceProcessesandResourceEcology,CollegeofGlobalChangeandEarthSystemScience,BeijingNormalUniversity,100875,Beijing,ChinabSchoolofGeography,BeijingNormalUniversity,100875,Beijing,ChinaabstractarticleinfoArticlehistory:Received28April2010Receivedinrevisedform13January2011Accepted19January2011Availableonline22February2011Keywords:Soilparticle-sizedistributionmapMappolygonlinkageSoiltypelinkageTaxotransferschemeWedevelopedamulti-layersoilparticle-sizedistributiondataset(sand,siltandclaycontent),basedonUSDA(UnitedStatesDepartmentofAgriculture)standardforregionallandandclimatemodellinginChina.
The1:1,000,000scalesoilmapofChinaand8595soilprolesfromtheSecondNationalSoilSurveyservedasthestartingpointforthiswork.
Wereclassiedtheinconsistentsoilprolesintothepropersoiltypeofthemapasmuchaspossiblebecausethesoilclassicationnamesofthemapunitsandproleswerenotquitethesame.
Thesand,siltandclaymapswerederivedusingthepolygonlinkagemethod,whichlinkedsoilprolesandmappolygonsconsideringthedistancebetweenthem,thesamplesizesoftheproles,andsoilclassicationinformation.
Forcomparison,asoiltypelinkagewasalsogeneratedbylinkingthemapunitsandsoilproleswiththesamesoiltype.
Thequalityofthederivedsoilfractionswasreliable.
Overall,themappolygonlinkageofferedbetterresultsthanthesoiltypelinkageortheHarmonizedWorldSoilDatabase.
Thedataset,witha1-kmresolution,canbeappliedtolandandclimatemodellingataregionalscale.
2011ElsevierB.
V.
Allrightsreserved.
1.
IntroductionParticle-sizedistribution(PSD)isabasicphysicalpropertyofsoilsthataffectsmanyimportantsoilattributes.
ThePSDsofsoilshavebeenwidelyusedforestimatingvarioussoilhydraulicproperties(AryaandParis,1981;HaverkampandParlange,1986;MinasnyandMcBratney,2007).
Thepercentageofsand,siltandclaywithinasoilproleisfrequentlyrequiredtodescribethephysicalprocessesinsoilbylandandclimatemodelsatregionalandglobalscales(Dickinsonetal.
,1993;Dai,2003;Sitchetal.
,2003;Gassmanetal.
,2007).
Despitetheimportanceofhavingpropersoilpropertiesforuseinthesemodels,thereisadearthofspatialinformationonthephysicalandhydraulicpropertiesofsoil,especiallyforChina.
Webbetal.
(1993)producedaglobaldatasetforthetopandbottomsoildepths,witha1°by1°spatialresolution,thatincludedthepercentagesofsand,siltandclayofindividualsoilhorizonsfor106soiltypesbycombiningtheSoilMapoftheWorldofFAO-UNESCO(FoodandAgricultureOrganizationoftheUnitedNations/UnitedNationsEducational,ScienticandCulturalOrganization)withtheWorldSoilDataFileofZobler(1986).
MillerandWhite(1998)developedtheCONUS-SOILdataset,whichincludesthesandandclayfractionsfor11standardlayers.
Reynoldsetal.
(2000)producedtheFAO-UNESCOglobal5-minutedistributionsofthesandandclayfractionsfortwolayers(0–30and30–100cm).
Batjes(2006)derivedsoilpropertiesforthe106soilunitsshownontheSoilMapoftheWorldforxeddepthintervalsof20cmuptoadepthof100cm.
Dijkshoornetal.
(2008)developedasoilandterraindatabaseatascaleof1:1,000,000forChinawith1430proles(ZhangandZhao,2008).
WhileFAOetal.
(2009)usedthe1:100,000scalesoilmapofChinaandsoilprolesfromtheWorldInventoryofSoilEmissionPotential(WISE),whichincludedonly61prolesfromChina,todeveloptheHarmonizedWorldSoilDatabase(HWSD),theyindicatedaneedformoresoilprolesfromChina.
Theexistingsoildatasetsarebasedonlimitedproledataandacoarseresolutionofspatialdata;therefore,theycannotsatisfytherequirementsofregionalmodellingforChina.
Thus,itremainscrucialtoupdateandexpandsoilPSDdatabasesthatarespecicallydesignedformodellingapplications.
Thegoalofthisstudyistodevelopapractical1-kmresolutiondatasetofparticle-sizedistributionofsoilforChinathatissuitableforregionallandandclimatemodelling.
2.
Dataandmethods2.
1.
DatasourceandpreparationThe1:1,000,000soilmapofChinawascompiledbytheInstituteofSoilScience,ChineseAcademyofSciences(Shietal.
,2004)basedontheresultsoftheSecondNationalSoilSurveyofChina.
ThismapisthemostdetailedsoilmapinChinaatthenationalscale.
ItisclassiedusingtheGeneticSoilClassicationofChina(GSCC),whichincludes12orders,61greatgroups,235sub-greatgroups,and909families.
Geoderma171-172(2012)85–91Correspondingauthorat:CollegeofGlobalChangeandEarthSystemScience,BeijingNormalUniversity,No.
19,XinjiekouwaiSt.
,Beijing100875,China.
Fax:+861062205274.
E-mailaddress:shanggv@hotmail.
com(W.
Shangguan).
0016-7061/$–seefrontmatter2011ElsevierB.
V.
Allrightsreserved.
doi:10.
1016/j.
geoderma.
2011.
01.
013ContentslistsavailableatScienceDirectGeodermajournalhomepage:www.
elsevier.
com/locate/geodermaThereare94,303mappolygonsinthemap,including85,257soilmappolygonsand9046non-soilmappolygons.
Morethanhalfofthesoilmappolygonsareatthesub-greatgrouplevel,andtheothersareatthegreatgrouporfamilylevel(Table1).
ThelatitudeandlongitudeofthecentresofthemappolygonswereextractedfromthecoverageleusingGIStools.
ThesoilproleswerefromtheChinesesoilproledatabase,whichwasalsoestablishedusingtheresultsoftheSecondNationalSoilSurveyofChinaconductedinthe1980s.
Itcontainsdatafor33,039soillayersrepresenting8979proles.
ThedatawerepublishedbytheNationalSoilSurveyOfce(1993a,b,1994,1995a,b,1996),provincialsoilsurveyofcesandthesoilsurveyofcesofsomeTibetancounties.
However,PSDdataarenotalwaysavailableforeachlayer.
Thus,thenumberofsamplesvarieswithsoiltypeanddepth.
Finesizefractionsweredeterminedusingthehydrometreorpipettemethod,whereascoarsesizefractionswereobtainedthroughsieving(NationalSoilSurveyOfce,1992).
Particle-sizefractiondatawereclassiedbyseveralschemesincludingtheISSS(InternationalSocietyofSoilScience)andKatschinski'sschemes.
Formodellingpurposes,theseparticle-sizedatawereconvertedtotheFAO-USDA(UnitedStatesDepartmentofAgriculture)System(ShangguanandDai,2009;ShangguanandDai,2010).
Thelatitudeandlongitudeofthesoilproleswerederivedatdifferentlevelsofspatialprecisionfromtheirgeographiclocationdescriptions.
Thespatialprecisionofprolelocationswasbrokendownintothreeclassications:A,BandC,whichhaderrorsbelow15km,between15kmand60km,andabove60km,respectively.
ThesoilclassicationsystemusedfortheaforementionedsoilprolesandsoilmapwastheGSCC.
However,therearesomeinconsistencies:therewereproleswithaclassicationataspecicsoiltypelevel(e.
g.
,soilfamily)thathadnocorrespondingmapunitofthesametypelevel,andviceversa;differentnameswereusedforthesamesoiltype.
Theinconsistentsoilproleswerereclassiedintothepropersoiltypefromthesoilmapatdifferentsoiltypelevel(i.
e.
,greatgroup,sub-greatgrouporfamily).
Basically,soiltypenamesweremodiedinlightoftheprincipleofapproximationofnaminganduseofbynamesforsoils.
Forexample,'plougheddiluviumsandythinmeadowsoil'wasmodiedas'sandythinmeadowsoil'.
Forsoilmapunitswithparentmaterialinformation,theparentmaterialofthecorrespondingsoilprolewasalsousedtomodifythesoiltypenames.
ThePSDdatawereinterpolatedto2and11standardlayersbyadepth-weightedmethodfortheirconvenienceofuseinlandandclimatemodels(Reynoldsetal.
,2000).
Manygrid-basedmodelsaredesignedasequal-compartmentlayers(Dickinsonetal.
,1993;Dai,2003).
However,thegreatrangeanddiversityofsoilprolelayerthicknessesmaketheminconvenienttouseinthesemodelswithoutadditionalanalyses.
The2layerswerethetopsoil(0–30cm)andsubsoil(30–100cm),andthe11layerswerethesameastheCONUS-SOILdatasetstandard,whichretainsabetterverticalvariation(MillerandWhite,1998).
Forbrevity,onlythe2-layerdatasetisshowninthispaperbecauseitiseasiertocomparewithotherdatasets.
SoilproleswithoutPSDdataorconsistentsoilclassicationinformationwereexcluded,leaving8595soilprolesfromwhichtoderivesand,siltandclaymapsbyalinkagemethod.
2.
2.
Methods2.
2.
1.
ExistinglinkagemethodandproblemsInpreviousstudies,thelinkagemethodhasusuallybeenaccomplishedbylinkingsoilmapunitsandprolesfollowingtheso-calledtaxotransferrules(Reynoldsetal.
,2000;Batjes,2003;FAOetal.
,2009).
Wecalledthismethodthesoiltypelinkage.
Soiltypelinkagegavesoilparametreestimatesbysoilunitsforeachsoillayer,usuallywithreferencetoasoil'stexturalclass.
Thevariationinsoilpropertiesacrossdifferentmappolygonsofthesamesoiltypethatactuallyexistedwasnotconsidered.
2.
2.
2.
Polygonlinkagemethod2.
2.
2.
1.
BasicIdea.
Inthisstudy,amethodoflinkingsoilprolestoindividualpolygonsinsteadofmapunitswasdeveloped.
Inordertopreservethespatialvariationinsoilpropertiesasmuchaspossible,twoaspectsotherthanthesoilclassicationinformationweretakenintoaccount(i.
e.
,thesamplesizesofprolesandthedistancesbetweensoilpolygonsandproles).
Thedistancewasusedtodeterminetheorderofpriorityofsoilprolestolinktoamappolygon.
Thelikelihoodoflinkagedecreasedasdistanceincreased,andthevariationofsoilpropertiesamongpolygonsofthesamesoiltypewasretained.
Thepossibleeffectsofregionalvariationinenvironmentalfactors(e.
g.
,climateandvegetation)werealsoimplicitlyconsidered.
Torepresentamappolygon,aminimumsamplesizeofsoilwasneeded.
Scholesetal.
(1995)insistedthatatleast30pedonspersoilunitwerenecessarytoprovideadequaterepresentationforthe106soilunitsoftheSoilMapoftheWorldata1:5,000,000scale.
FortheSoilsandTerrainDatabase(SOTER),eachsoilcomponentofamapunitwascharacterisedbyatypicalsoilprole(Batjesetal.
,2007).
InordertollinthegapsintheprimarySOTERdatabase,Batjes(2003)performedataxotransferschemeusingthemedianofmorethan5prolesfortheconsideredcombinationofFAOsoilunit(orsoilgrouping),attribute,depthzoneandsoiltextureclass.
Mapunitswithhighersoiltypelevelsneedmoresamplestocoverthevariationinsoilproperties.
Inthisstudy,weaimedforatleast40prolesforagreatgroup,10prolesforasub-greatgroupand3prolesforasoilfamily.
2.
2.
2.
2.
Linkageprocess.
First,asareferenceofpolygonlinkage,theEulerdistancebetweenamappolygonandasoilprolewiththesamesoiltypewascalculated.
Soilproleswithoutsufcientprecisionfortheirlocationwerenotinvolvedintheselectionoflinkages.
Themappolygonlinkageprocesswasperformedasfollows:1.
Soilprolesofthesamesoiltypeforeachmappolygonweresearchedata15-kmradius.
Weassumedthatallprolesinthisrange(aboutthecountysize)shouldbeusedtorepresentamappolygon.
Inaddition,theprolesinthisradiuswerelikelytobewithinornearthelinkedmappolygon,astheaveragesizeofmappolygonswasabout10kmby10km.
Iftherewereenoughsoilprolesofthesamesoiltypeinthisradius,theseproleswerelinkedtothemappolygon.
Otherwise,wecontinuedtostep2.
SoilproleswithspatialprecisionsBandCwerenotinvolvedinthisstep.
2.
ThesearchradiuswasenlargeduntilitwasgreaterthanthewholesoilmapofChinaorthetargetnumberofsoilproleswasreached.
Ifthetargetnumberofsoilproleswasreached,thesesoilproleswerelinkedtothemappolygonandthenalradiuswasrecorded.
Ifthesearchresultedininsufcientsoilproles,thesesoilproleswerealsolinkedtothemappolygonandmarkedasinsufcientlyrepresented.
Iftherewasnosoilproleinthewholesoilmap,weproceededtostep3.
3.
Thegroupinglevelofsoiltypewasexpandedandthesearchwasrestartedatstep1.
Forexample,iftherewasnoproleforamappolygonatthefamilylevel,thesearchwasrestartedatthesub-greatgrouplevel.
Thelinkagestartedatthelowestsoiltypelevelofamappolygonandcontinuedupwardstothegreatgrouplevel,withdifferentlinkagesstoppingatdifferentlevels.
Table1Numbersofdifferentsoiltypeleveloflinkageatdifferentsoiltypelevelofmappolygons.
SoiltypelevelofmappolygonsSoiltypeleveloflinkageFamilySub-greatgroupGreatgroupSubtotalFamily20,1614575524,741Sub-greatgroup–42,81548943,304Greatgroup––13,81213,812Subtotal20,16147,39014,30681,85786W.
Shangguanetal.
/Geoderma171-172(2012)85–912.
2.
2.
3.
Obtainingarepresentativevalue.
Themedians,means,ranges,variancesandsamplesizesforthesand,siltandclaycontentsofthelinkedsoilproleswerecalculatedbothforthetopsoil(0–30cm)andsubsoil(30–100cm).
Themapsofsand,siltandclaycontentwerederivedbyusingthemedianvalueforeachsoilpolygon(Batjes,2006)becausetheinuenceofextremevaluesispartiallyignoredcomparedtoameanvalue.
ThePSDcomputedbasedonmediansrarelysummedupto100%.
Toguaranteethatthesumofthreefractionstotalled100%,thefollowingprocesswasadopted.
First,themedianofeachfractionforlinkedsoilproleswascalculated.
Then,thesumofsquareddeviation(SSD)ofthemedianswascalculatedforeachlinkedprolebasedonthefollowingformula:SSDi=saisam2+siisim2+cliclm21wheresai,siiandcliarethesand,siltandclayfractionoftheithlinkedprole,respectively,andsam,simandclmarethemediansofthesand,siltandclayfractionsforthelinkedsoilproles,respectively.
Finally,theprolewiththeminimumSSDwasusedtorepresentthesoilpolygon.
Forgrid-basedmodelapplications,thevector-formatdataweresubsequentlyrasterisedtospacedgridsataresolutionofabout1-km(30arcsecondsby30arcseconds)forsand,siltandclay.
Therewerenon-soilmappolygons(organicmaterials,water,rocksorother)andlayerscontainingbedrock.
Asaresult,thesumofthecomputedsand,siltandclayfractionswasoftenlessthan100%whenrasterisationwasdone.
Thesand,siltandclayfractionswerenormalisedto100%(beforerounding)ifthesumofthefractionswaslessthan100%andgreaterthan50%.
Otherwise,thefractionsweresettozero(Thisnormalisationmaycausesomefalseinformationtobeincluded;MillerandWhite,1998).
2.
3.
ValidationandcomparisonToevaluateandvalidatetheresultsofthelinkagemethod,anindependentdatasetwasused.
Thedatawerecollectedfromthreeareasin2008and2009.
Therewere168,163and58samplesfromtheFig.
1.
Sand,silt,andclayfractionsofChina.
(a)Sandfractionofthetopsoil(0–30cm).
(b)Siltfractionofthetopsoil(0–30cm).
(c)Clayfractionofthetopsoil(0–30cm).
(d)Sandfractionofthesubsoil(30–100cm).
(e)Siltfractionofthesubsoil(30–100cm).
(f)Clayfractionofthesubsoil(30–100cm).
87W.
Shangguanetal.
/Geoderma171-172(2012)85–91BingxiancountyofHeilongjianprovince(3834km2),AnsaicountyofShaanxiprovince(3607km2,includingpartsoftheneighbouringcounties)andZitongcountyofSichuanprovince(1435km2),respec-tively.
Thesamplesweretakenasa5-kmgridforthetopsoillayer(0–20cm).
Thenesizefractionsweredeterminedusingthehydrometremethod,whereasthecoarsesizefractionswereobtainedthroughsieving.
Bingxianisdominatedbyblacksoil(whichisablack-colouredsoilcontainingahighpercentageofhumusandhighpercentagesofphosphoricacids,phosphorusandammonia,correspondingtoPhaeo-zemsinWorldReferenceBaseforsoilresources(WRB)),meadowsoil(whichcontainsahighpercentageofhumuswithahighgroundwaterlevelandmeadowvegetation,correspondingtoCambisolsinWRB)anddarkbrownsoil(whichisadarkbrown-colouredsoilcontainingahighpercentageofhumuswithvegetationofconiferousandbroad-leavedmixedforest,correspondingtoCambisolsinWRB).
Ansaiisdominatedbyloessialsoil(whichhasapparentcharacteristicsofparentmaterialofloess,correspondingtoCambisolsinWRB),andZitongisdominatedbypurplishsoil(whichisdevelopedfrompurplishshaleandsandstone,andattheearlystageofeluviations,correspondingtoCambisolsinWRB).
ThoughasoilgreatgroupinGSCCcouldbeinterpretedintoseveralWRBsoilgroups(Shietal.
,2010),onlythedominantoneweregivenhere.
Thecross-referencewasalsodevelopedtorelateGSCCwithSoilTaxonomyofUSandChineseSoilTaxonomy(Shietal.
,2006a,b).
Thequalityofthelinkagewasevaluatedbasedonthesearchradius,soiltypelevelofthelinkageandsamplesize.
Ifthesearchstoppedatasmallradius,itisimpliedthatthelinkedprolesareclosetothemappolygonsandoffergoodestimates.
Ifthesoiltypeleveloflinkageislow(suchassoilfamily),thevariationinsoilpropertiesislowerthanathigherlevelsofsoiltype.
Ifthetargetforsamplesizeisreached,thepolygoncanbeconsideredwellrepresented.
Forcomparison,thesoiltypelinkagemethodwasalsoperformed,followingthemethodsofpreviousstudiesinwhichmapunitsandproleswiththesamesoilclassicationinformationwerelinked(Reynoldsetal.
,2000;Batjes,2003;FAOetal.
,2009).
Theresultsderivedthroughthesoiltypeandpolygonlinkagemethodswerecomparedwiththeindependentsamplesusingmeanerror(ME)androotmeansquareerror(RMSE).
TheHarmonizedWorldSoilDatabase(HWSD),whichwasderivedbylinkingsoilmapunitsandprolesfromWISE,wasalsocomparedwithourresults.
3.
ResultsanddiscussionFig.
1showsthesand,siltandclayfractionslinkedbymappolygons.
ThefractionmapsdisplaysoilPSDdistributionforChinaingreatdetail.
Generally,northandwestChinahavehighsandfractionsandlowclayfractions,especiallyinthedesertarea,whiletheoppositewasobservedinsouthChina.
ThisisexpectedduetothephysicalandchemicalweatheringprocessesindifferentpartsofChina.
Thenon-soilmapunitswereassignedzerovaluesforallthreefractions.
ThequalityofthederivedPSDdatasetwasassessedbasedonthelevelofsoiltypelinkage,samplesizeandsearchradius.
ThelinkagelevelsofmappolygonsareshowninTable1.
Mostofthesoilmappolygonswerelinkedatthesamesoiltypeleveltheybelongto,whichindicatesthatmostofsoiltypeshadcorrespondingsoilproles.
Thelinkagelevelwasrecordedforeachsoilmappolygonforfuturereference.
LowerlevellinkageshadbetterestimatesforPSDs.
Thetargetsamplesizewasachieved,exceptforin9.
2%,7.
8%and1.
2%ofthelinkagesatthelevelsofsoilfamily,soilsub-groupandsoilgroup,respectively.
Themappolygonsthatwerenotlinkedatthesamesoiltypelevelthattheybelongtoordidnotreachthetargetsamplesizeneedmoreprolesamplesexaminedinthefuture.
Ontheotherhand,thesamplesizewasmaintainedatthetargetnumbertodescribethevarietyofdifferentmappolygonsofthesamesoiltype,thoughScholesetal.
(1995)chosenottoexcludeadditionalsoilprolesonthebasisthataparticularsoiltypewasalreadywellrepresented.
Fig.
2showsthecountsofdifferentlinkageradiusbetweenmappolygonsandsoilproles.
Themedianvaluewasabout146kmandthe75thpercentilewasabout520km,whichindicatesthatmostlinkagehappensattheclimatezonescale.
Althoughnaturalsimilarityandvarietywasmainlyconsideredwithinthecontextofthesoilmapitselfinpreviousstudies(Webbetal.
,1993;Reynoldsetal.
,2000;Batjes,2006),thedistance-Fig.
2.
Thedistributionoflinkageradiusbetweenmappolygonsandsoilproles.
Fig.
3.
Differencesinsoilfractioncontentslinkedbymappolygonsandlinkedbysoiltypes.
(a)Sandfractionofthetopsoil(0–30cm).
(b)Clayfractionofthetopsoil(0–30cm).
88W.
Shangguanetal.
/Geoderma171-172(2012)85–91basedlinkingmethod,whichconsidersthedistancebetweensoilpolygonsandproles,betterpresentstheseproperties.
Fig.
3wasobtainedbysubtractingthesandandclaycontentderivedbysoiltypesfromthosederivedbymappolygons.
Thecontentsofthesesoilfractionsweredifferentinmostofthemappolygons.
Forthesoiltypelinkage,thesandcontentinthenorthandsoutheastandtheclaycontentinthesouthwestwereunderestimatedcomparedtothoseofthemappolygonlinkage.
Thedifferencesconrmthatitisbettertoderivesoilpropertiesthroughalinkagemethodthatconsidersthedistancebetweenprolesandmappolygonstopresentthevariationinsoilpropertiesofdifferentpolygonswiththesamesoiltype.
Whendetaileddatasetsareavailable,themappolygonlinkagemethodoffersmorespatialinforma-tionforsoilfractions.
Iftherearenotenoughsoilproles,themappolygonlinkageandsoiltypelinkagewillnotappeartobesignicantlydifferent,asthepolygonlinkagewillnotstopuntilitisoutsideofthewholesoilmapareainstep2.
Itisalsoreasonablethatthesoiltextureforthesamesoiltypevarieswithinacertainrangeindifferentlocations,whichusuallymeansthereisadifferenceinsoilformationfactors,i.
e.
,climate,organisms(includinghumans),relief,parentmaterialandtime(Jenny,1941),particularlyforsoilsgroupedtogetheratahighclassicationlevel.
Table2showstheMEandRMSEofsoilfractioncontentsforthethreedatasets:HWSD,thepolygonlinkagedatasetandthetypelinkagedataset.
Overall,thepolygonlinkagemethodgavethemostaccurateestimationandHWSDgavetheleastaccurateestimation,withtheexceptionofthepolygonlinkageperformingslightlyworsethantheothertwodatasetsforestimatingclaycontents.
Withinthelimitsofourdata,thereweresomedifferencesinthePSDsofthetopsoildepthofsamplescollectedforvalidationandthePSDmaps.
IntheAnsaiandBingxianareas,allofthedatasetsoverestimatedsandandclaycontentsandunderestimatedsiltcontents.
However,intheZitongarea,thepolygonlinkageandtypelinkagemethodsoverestimatedsiltcontentsandunderestimatedsandcontents,whiletheoppositehappenedwithHWSD.
Thisindicatesthattheperformanceofthesedatasetsvarieswithsoiltype,astheseareashavedifferentsoiltypes.
Inallareas,thepolygonlinkageestimateshadthelowestRMSEsforsandandsiltcontents,butintheAnsaiandBingxianareas,thepolygonlinkagemethoddidnotperformthebestofthethreemethods.
Thesourcesofuncertaintyinthelinkagemethodshavebeendiscussedinpreviousstudies(Batjes,2002;Batjes,2006).
Errorsinspatialdataaremuchmoreimportantthanthoseinsoilanalyticalmethodsbecauseofthepurityofsoilmapunits,whichislikelytobearound50to65%(Landon,1991).
The1:1,000,000scalesoilmapofChinawascompiledthroughthecartographicgeneralisationof1,500,000scalemaps.
However,thisresultsinthelossofsoiltypemakeupinformation,leavingonlyasinglesoiltypepermappolygon,whichdegradesthequalityofthespatialdata.
Theimpurityinsoilmapunits,whichisnottakenintoaccountinthelinkagemethods,cancausesignicanterrorsinestimatingsoilfractions,asothersoiltypeswithinamapunitormappolygonmayhavequitedifferenttexturesthantheonetowhichitislinked.
Thelinkagemethodmaybeimprovedbyaddingallsoilproleswithinamappolygontothelinkedprolestodeterminetherepresentativevaluesofsoilproperties.
However,becausethereisonlylessthan0.
1soilprolepermappolygoninthedatabaseforChina,anyimprovementwouldbeverysmall.
Inadditiontothetwosourcesofuncertaintymentionedabove,theaccuracyofthedistancesbetweenmappolygonsandproles,thesoilclassicationsystemandthelinkagemethoditselfcanalsocarryuncertainty.
Thecoordinatesofsoilproleswerenotveryaccuratebecausetheywerederivedfromthelocationdescription.
Thecentre,ratherthantheboundary,ofamappolygonwasusedtocalculatethedistancesbetweenmappolygonsandproles.
Therefore,thedistanceshadsomeassociatederror.
Inthepolygonlinkage,distancewasusedtodeterminewhetheraproleshouldbelinkedtoamappolygon.
Sinceweabandonedprolesthatdidnothavesufcientprecisionoflocationanddidnotweightanyofthedistances,theeffectofdistanceerrorsonthelinkingresultswasrathersmall.
TheGSCCsystemhassomeshortcomings(Gong,1999).
Itisbasedonthesoilgenetichypothesis,whichmayresultinthesamesoilbeingclassiedasdifferentsoiltypes.
Forexample,albicsoilswereclassiedaspodzolicsoilsbecausethealbicprocessandpodzolicprocesseswerenotyetdistinguishedinthe1950s.
Inaddition,theGSCCemphasisestheimportanceofclimateandvegetationwhileignoringthetimefactor.
Therefore,itmayendupconfusingsoil-formingprocessesthathavealreadyhappenedwiththosewhichhavenotyetoccurred.
Forexample,underextremeconditions,itmayevenclassifyapurplishsoilasayellowsoil(whichhasanintensiveeluviationwithhighcontentofgoethite,correspondingtoCambisolsinWRB).
TheGSCCemphasisestheCentralConcept,whichstatesthat,whilesoiltypecanbeveryclear,theboundariesbetweentypesmaybeunclear,makingsomesoilshardtoclassifyasaspecicsoil.
TheGSCCalsolacksquantitativeindices,whichcausesitsinformationsystemtobedifculttobuild.
Becauseoftheshortcomingsmentionedabove,itishardtoavoiderrorsincompilingsoilmapsandclassifyingsoilproles.
Inaddition,thesoilsurveyemployedabottom-upprocedurestartingfromthecountyortownlevel,andasaresult,inconsistenciesareinevitableduetodifferencesinthepersonaljudgmentsofthedatacollectors.
Aspreviouslymentioned,soiltypenamesofsoilprolesweremodiedtobeconsistentwiththesoilmapatdifferentlevelsofsoiltype(notalwaysthelowestlevelofsoiltype),whichcanalsocausesomeuncertainty.
Whilethelinkagemethodusedasinglesoilcontentvaluetorepresentapolygonoramapunit,thetextureofsoilcanvaryspatially(sometimessignicantly)withinaspecicpolygonormapunit.
Inthisstudy,thepolygonlinkagemethodwasusedtotakeintoaccounttheinter-polygonalvariationwithinamapunit.
Thiswasnotconsideredbythesoiltypelinkagemethod.
Thepolygonlinkagemethodcannotrepresentthespatialvariationwithinapolygon,eventhoughitsstatisticalvariationcanbegivenbythevarianceandrangeofthelinkedsamples.
Analternativemethodthattakesintra-polygonalvariationintoaccountistheBayesianMaximumEntropymethod(BME),whichusessoiltextureclassmapsasinputdata.
(D'OrandBogaert,2003).
InChina,soiltextureclassmapofhighprecisiondoesnotexist;however,thelinkagemethodcancreateone.
TheunderlyingassumptionofBMEisthecontinuouschangeofsoilpropertiesinsidethepolygon,whichoftenappliesatanescale(asinthecaseind'OrandBogaert,2003).
Atacoarsescale,suchasthe1:1,000,000scaleinourstudy,thisassumptionismorelikelytobewrong.
Sincetheaveragepolygonsizeisabout100km2,thevaluesofsoilfractionsarenotlikelytochangegraduallyfromthecentretotheboundary,butratherthroughaseriesofupsanddownsalongtheway.
Theprocessofchoosingthelinkagebetweensoilmappolygonsandproleswassubjective,asthetargetsamplesizeforasoiltypelevelandthesearchradiusatstep1ofthelinkageprocesswastosomeextentarbitrarilydecided.
Inaddition,thelinkagemethoddoesnotapplytothenon-soilmapunits,suchascityareasthatdohavesoil,whichweresettoavalueofzeroforallsoilfractions.
Table2Accuracyofsoilfractioncontentsfromthreedatasetsinthreeareas.
AreaSourceaSandSiltClayMERMSEMERMSEMERMSEAnsaiPolygon1.
113.
67.
212.
86.
17.
1Type13.
617.
014.
016.
30.
32.
8HWSD16.
918.
917.
720.
12.
14.
6BingxianPolygon1.
911.
08.
212.
06.
310.
9Type7.
817.
713.
316.
25.
513.
2HWSD10.
518.
119.
322.
28.
910.
2ZitongPolygon2.
213.
05.
612.
03.
47.
1Type11.
018.
07.
613.
23.
47.
5HWSD12.
818.
58.
013.
34.
87.
8TotalPolygon0.
512.
75.
412.
34.
89.
7Type7.
017.
310.
015.
62.
99.
1HWSD13.
118.
416.
620.
04.
18.
3a"Polygon"wasderivedbylinkingsoilmappolygonsandproles.
"Type"wasderivedbylinkingsoilmapunitsandproles.
"HWSD"istheHarmonizedWorldSoilDatabase.
89W.
Shangguanetal.
/Geoderma171-172(2012)85–91Thepolygonlinkageconsidersthesoiltypeanddistancebetweensoilpolygonsandprolesandindirectlytakesintoaccountenvironment-relatedfactors.
Asthespatialvariationofclimateandvegetationisrelativelysmall,ourdistance-basedmethodmaybesufcientbecausemostofthelinkagehappensattheclimatezonescale.
IntheSecondNationalSoilSurveyofChina,topographicmaps,orairphotos,wereusedasthebasemapsofsoiltypemaps,withgeologicalmapsasareference.
Tosomeextent,thefactorsoftopography,landuseandparentmaterialwereimplicitlyconsid-ered.
However,thesefactorsvarygreatly,soadistance-basedlinkageisnotabletocapturethemwell.
Inthefuture,itwillbenecessarytoexplicitlyconsidertheseenvironmentallyrelatedfactors.
Inthiscontext,thepolygonlinkageisatypicalexampleofthescorpanparadigmofquantitativeempiricaldigitalsoilmapping(McBratneyetal.
,2003).
Zhaoetal.
(2006)developedapedologicalknowledge-basedmethod,whichconsiderssoilclassicationinformationandthelocationsofproles.
Theproleswithinacountywerealllinkedtomappolygonsinthesamecountyinthatstudy,sincetheSecondNationalSoilSurveyofChinawasimplementedfromthecountylevel.
ThoughZhaoetal.
(2006)indirectlyconsideredspatiallocation,theactualspatialpatternofsoilpropertieswasnotconnedbyadministrativedivisionboundaries.
Itisbettertotakelocationintoaccountthroughthedistancebetweenprolesandmappolygons,likeinourstudy.
4.
ConclusionsAsoilPSDdatasetwith1-kmresolutionwasdevelopedforitsapplicationinlandandclimatemodellingbyusingthemostdetailedsoilmapofChinaatthenationalscaleandalargesoilproledatabase.
ThepolygonlinkagemethodprovidesmoreinformationaboutthedistributionofsoilPSDsthanthesoiltypelinkagemethod.
Theoverallassumptionsarethatasoilmappolygoncanberepresentedbyaminimumsamplesizeofsoilandthatsoilfractionsvaryduetothesoiltypeandlocation.
Accordingtothesoiltypeleveloflinkage,samplesizeandsearchradius,thequalityofthedatawasreliable.
Overall,themappolygonlinkageofferedbetterresultsthanthesoiltypelinkageorHarmonizedWorldSoilDatabase.
However,weneedtoputeffortintoimprovingthedataqualityandtheproductaccuracy.
Thedatasetisavailableforfreedownloadfromhttp://globalechange.
bnu.
edu.
cn.
Futureeffortswillbemadetoimprovethequalityofthedataset.
Thislinkagemethodmaybeusedinfutureworktoderiveothersoilproperties,suchasrockfragmentcontent,soildepth,andsoilcarbonandnitrogencontents.
Itisalsonecessarytoconsiderenvironment-relatedfactorsdirectlyunderthescorpanframeworktoimprovethepredictionofsoilproperties(McBratneyetal.
,2003).
AcknowledgementsThisworkwassupportedbyfundingagenciesincluding,theNSFCunderGrant40775041,theR&DSpecialFundforNonprotIndustry(Meteorology,GYHY200706005andGYHY200706025)andMOSTNo.
2010CB951802.
WewouldliketoacknowledgeA-XingZhuandLizongWufortheirassistanceandhelpfuldiscussions.
Wewouldalsoliketothanktwoanonymousreviewersfortheirthoroughandconstructivereviews.
AgroupofstudentsatBeijingNormalUniversityhelpedwiththedatacollection.
TheauthorsthankPeggyChenforhelpcorrectingtheEnglishwriting.
ReferencesArya,L.
M.
,Paris,J.
F.
,1981.
Aphysicoempiricalmodeltopredictthesoilmoisturecharacteristicfromparticle-sizedistributionandbulkdensitydata.
SoilSci.
Soc.
Am.
J.
45,1023–1030.
Batjes,N.
H.
,2002.
Soilparameterestimatesforsoiltypesoftheworldforuseinglobalandregionalmodelling(version2.
1).
availablethroughISRICRe-port2002/02c.
InternationalFoodPolicyResearchInstitute(IFPRI)andInterna-tionalSoilReferenceandInformationCentre(ISRIC,Wageningen.
http://www.
isric.
org.
Batjes,N.
H.
,2003.
Ataxotransferrule-basedapproachforllinggapsinmeasureddatainprimarySOTERdatabases.
ISRICReport2003/03.
ISRC-WorldSoilInformation,Wageningen.
Batjes,N.
H.
,2006.
ISRC-WISEderivedsoilpropertiesona5by5arc-minutesglobalgrid(version1.
0).
availablethroughISRCReport2006/02.
ISRIC-WorldSoilInformation,Wageningen.
http://www.
isric.
org(withdataset).
Batjes,N.
H.
,Al-Adamat,R.
,Bhattacharyya,T.
,Bernoux,M.
,Cerri,C.
E.
P.
,Gicheru,P.
,Kamoni,P.
,Milne,E.
,Pal,D.
K.
,Rawajh,Z.
,2007.
PreparationofconsistentsoildatasetsforSOCmodellingpurposes:secondarySOTERdatasetsforfourcasestudyareas.
Agric.
Ecosyst.
Environ.
122,26–34.
Dai,Y.
,2003.
Thecommonlandmodel.
Bull.
Amer.
Meteor.
Soc.
84,1013–1023.
Dickinson,R.
E.
,Henderson–Sellers,A.
,Kennedy,P.
J.
,Wilson,M.
F.
,1993.
Biosphereatmospheretransferscheme(BATS)version1eascoupledforCommunityClimateModel.
NCARTech.
NoteNCAR/TN–378+STR,p.
72.
Dijkshoorn,J.
A.
,vanEngelen,V.
W.
P.
,Huting,J.
R.
M.
,2008.
SoilandlandformpropertiesforLADApartnercountries(Argentina,China,Cuba,Senegal,TheGambia,SouthAfrica,Tunisia).
ISRICreport2008/06,GLADAreport2008/03.
ISRIC–WorldSoilInformation,FAO,Wageningen(23ppwithdataset).
D'Or,D.
,Bogaert,P.
,2003.
Continuous-valuedmapreconstructionwiththeBayesianMaximumEntropy.
Geoderma112,169–178.
FAO,IIASA,ISRIC,ISS-CAS,JRC,2009.
HarmonizedWorldSoilDatabase(version1.
1).
FAO,Rome,ItalyandIIASA,Laxenburg,Austia.
Gassman,P.
W.
,Reyes,M.
R.
,Green,C.
H.
,Arnold,J.
G.
,2007.
Thesoilandwaterassessmenttool:historicaldevelopment,applicationsandfutureresearchdirec-tions.
Trans.
Am.
Soc.
Agric.
Biol.
Eng.
50,1211–1250.
Gong,Z.
,1999.
Chinesesoiltaxonomy:theory,methodologyandpractice.
SciencePressofChina,Beijing.
26–28.
Haverkamp,R.
,Parlange,J.
Y.
,1986.
Predictingthewater-retentioncurvefromaparticle-sizedistribution:1.
Sandysoilswithoutorganicmatter.
SoilSci.
142,325–339.
Jenny,H.
,1941.
FactorsofSoilFormation,ASystemofQuantitativePedology.
McGraw-Hill,NewYork.
Landon,J.
R.
,1991.
BookerTropicalSoilManual.
LongmanScientic&Technical,NewYork.
McBratney,A.
B.
,MendoncaSantos,M.
L.
,Minasny,B.
,2003.
Ondigitalsoilmapping.
Geoderma117,3–52.
Miller,D.
A.
,White,R.
A.
,1998.
AconterminousUnitedStatesmultilayersoilcharacteristicsdatasetforregionalclimateandhydrologymodeling.
AvailableonlineatEarthInteractions2.
http://EarthInteractions.
org.
Minasny,B.
,McBratney,A.
B.
,2007.
Estimatingthewaterretentionshapeparameterfromsand,claycontent.
SoilSci.
Soc.
Am.
J.
71,1105–1110.
NationalSoilSurveyOfce,1992.
Chinasoilsurveytechnology.
ChinaAgriculturePress,Beijing,China.
inChinese.
NationalSoilSurveyOfce,1993a.
Chinasoilspeciesdescription,Vol.
1.
ChinaAgriculturePress,Beijing,China.
inChinese.
NationalSoilSurveyOfce,1993b.
Chinasoilspeciesdescription,Vol.
2.
ChinaAgriculturePress,Beijing,China.
inChinese.
NationalSoilSurveyOfce,1994.
Chinasoilspeciesdescription,Vol.
3.
ChinaAgriculturePress,Beijing,China.
inChinese.
NationalSoilSurveyOfce,1995a.
Chinasoilspeciesdescription,Vol.
4.
ChinaAgriculturePress,Beijing,China.
inChinese.
NationalSoilSurveyOfce,1995b.
Chinasoilspeciesdescription,Vol.
5.
ChinaAgriculturePress,Beijing,China.
inChinese.
NationalSoilSurveyOfce,1996.
Chinasoilspeciesdescription,Vol.
6.
ChinaAgriculturePress,Beijing,China.
inChinese.
Reynolds,C.
A.
,Jackson,T.
J.
,Rawls,W.
J.
,2000.
Estimatingsoilwater-holdingcapacitiesbylinkingtheFoodAgricultureOrganizationsoilmapoftheworldwithglobalpedondatabases,continuouspedotransferfunctions.
WaterResour.
Res.
36,3653–3662.
Scholes,R.
J.
,Skole,D.
,Ingram,J.
S.
,1995.
AGlobalDatabaseofSoilProperties:ProposalforImplementation.
ReportoftheGlobalSoilsTaskGroup,IGBP-DIS.
35pp.
Shangguan,W.
,Dai,Y.
,2009.
InvestigationofdifferentmodelstodescribesoilParticle-sizedistributionforsparseexperimentaldata.
J.
BeijingNorm.
Univ.
Nat.
Sci.
45,279–283.
Shangguan,W.
,Dai,Y.
,2010.
Aninvestigationofsoilparticle-sizedistributionmodelsfortheconversionofsoiltextureclassicationfromISSS,Katschinski'stoUSDASystem.
InpreparationShi,X.
Z.
,Yu,D.
S.
,Warner,E.
D.
,Pan,X.
,Petersen,G.
W.
,Gong,Z.
G.
,Weindorf,D.
C.
,2004.
Soildatabaseof1:1,000,000digitalsoilsurveyandreferencesystemoftheChineseGeneticSoilClassicationSystem.
SoilSurv.
Horiz.
45,129–136.
Shi,X.
Z.
,Yu,D.
S.
,Warner,E.
D.
,Sun,W.
X.
,Petersen,G.
W.
,Gong,Z.
T.
,2006a.
Cross-referencesystemfortranslatingbetweenGeneticSoilClassicationofChinaandSoilTaxonomy.
SoilSci.
Soc.
Am.
J.
70,78–83.
Shi,X.
Z.
,Yu,D.
S.
,Xu,S.
X.
,Warner,E.
D.
,Wang,H.
J.
,Sun,W.
X.
,Zhao,Y.
C.
,Gong,Z.
T.
,2010.
Cross-referenceforrelatingGeneticSoilClassicationofChinawithWRBatdifferentscales.
Geoderma155,344–350.
Shi,X.
Z.
,Yu,D.
S.
,Yang,G.
X.
,Wang,H.
J.
,Sun,W.
X.
,Du,G.
H.
,Gong,Z.
T.
,2006b.
Cross-referencebenchmarksfortranslatingthegeneticsoilclassicationofChinaintotheChineseSoilTaxonomy.
Pedosphere16,147–153.
Sitch,S.
,Smith,B.
,Prentice,I.
C.
,Arneth,A.
,Bondeau,A.
,Cramer,W.
,Kaplan,J.
O.
,Levis,S.
,Lucht,W.
,Sykes,M.
T.
,Thonicke,K.
,Venevsky,S.
,2003.
Evaluationofecosystem90W.
Shangguanetal.
/Geoderma171-172(2012)85–91dynamics,plantgeographyandterrestrialcarboncyclingintheLPJdynamicglobalvegetationmodel.
Glob.
ChangeBiol.
9,161–185.
Webb,R.
S.
,Rosenzweig,C.
E.
,Levine,E.
R.
,1993.
Specifyinglandsurfacecharacteristicsingeneralcirculationmodels:soilproledatasetandderivedwater-holdingcapacities.
Glob.
Biogeo.
Cyc.
7,97–108.
Zhang,G.
,Zhao,Y.
,2008.
SOTERdatabaseforChina,scale1:1million.
InstituteofSoilScience,ChineseInstituteofSoilScience,Nanjing.
Zhao,Y.
,Shi,X.
,Weindorf,D.
C.
,Yu,D.
,Sun,W.
,Wang,H.
,2006.
MapscaleeffectsonsoilorganiccarbonstockestimationinNorthChina.
SoilSci.
Soc.
Am.
J.
70,1377–1386.
Zobler,L.
,1986.
AWorldSoilFileforGlobalClimateModelling.
NASATechnicalMemorandum#87802.
NASAGoddardInstituteforSpaceStudies,NewYork,U.
S.
A.
91W.
Shangguanetal.
/Geoderma171-172(2012)85–91