![P and Fe concentrations in modern soils' B horizons, binned by soil order. Red lines are median values, blue bins are 25th and 75th quartile, and red crosses are beyond the 75th quartile; the horizontal dashed gray line is the crustal average (870 ppm P and 3.5% Fe) [77]. (A) P concentrations, showing that younger soils (Entisols, Inceptisols) tend to have higher P than older soils (Alfisols, Ultisols). Oxisols have high ranges of P, but a much smaller bin size than other orders (n = 8). Andisols have high inherited P from volcanic parent materials. (B) Fe concentrations, showing consistent Fe concentrations between ca. 1 and 10 wt% with the exception of Oxisols, which are defined by their high rates of Fe oxide accumulation. Bin sizes: Entisols (44), Inceptisols (75), Alfisols (42), Ultisols (6), Oxisols (8), Aridisols (76), Mollisols (60), Andisols (15), Spodosols (9), Gelisols (30), Unknown (39).](https://www.researchgate.net/profile/Rebecca-Dzombak/publication/347242611/figure/fig1/AS:969020979306503@1608044013717/P-and-Fe-concentrations-in-modern-soils-B-horizons-binned-by-soil-order-Red-lines-are_Q320.jpg)
![P concentrations in all horizons, binned by vegetation type. Dashed gray line in all is the crustal average P (870 ppm) [77]. Arrows indicate off-plot values. (A) P in the Top 5 cm, where Barren (no vegetation) and Altered soils have the lowest range of P concentrations. (B) P in A horizons, with similar distributions to Top 5 cm. (C) P in B horizons, with much smaller bin sizes than other horizons. (D) P in C horizons. Bin sizes for (A,B,D): Barren (68), Altered (209), Forest (1252), Herbaceous (815), Developed (1568), Shrubland (945). Bin sizes for (C): Forest (43), Grassland/Herbaceous (19), Shrubland (50), Barren (5), Cultivated (59), Unknown (74).](https://www.researchgate.net/profile/Rebecca-Dzombak/publication/347242611/figure/fig2/AS:969020979310596@1608044013776/P-concentrations-in-all-horizons-binned-by-vegetation-type-Dashed-gray-line-in-all-is_Q320.jpg)
![Fe concentrations in all horizons, binned by vegetation type. Dashed gray line in all is the crustal average (3.5%) [77]. Arrows indicate off-plot values. (A) Fe in the Top 5 cm, where Barren and Altered soils have the lowest maximum Fe concentrations. (B) Fe in A horizons, with similar trends to Top 5 cm. (C) Fe in B horizons. There were no data for Barren B horizons' Fe. (D) Fe in C horizons. Bin sizes for (A,B,D): Barren (68), Altered (209), Forest (1252), Herbaceous (815), Developed (1568), Shrubland (945). Bin sizes for (C): Forest (43), Grassland/Herbaceous (19), Shrubland (50), Barren (0), Cultivated (59), Unknown (74).](https://www.researchgate.net/profile/Rebecca-Dzombak/publication/347242611/figure/fig3/AS:969020979281921@1608044013821/Fe-concentrations-in-all-horizons-binned-by-vegetation-type-Dashed-gray-line-in-all-is_Q320.jpg)
![(A) Latitudinal distribution of soils used in this study, where smaller bins represent the USGS dataset [64] and the wider bins are our soils. The range of latitudes covered here is 7° to 73° N, but most of the samples fall between 20 and 50° N. (B) Latitudinal trends in Chemical Index of Alteration (CIA). (C) Latitudinal trends in clay content. (D) Correlation between CIA and clay content (R 2 : 0.05, <0.01, and 0.19 for A, B, and C horizons, respectively). Overall, these results support common assumptions about the relationships between latitude and weathering.](https://www.researchgate.net/profile/Rebecca-Dzombak/publication/347242611/figure/fig4/AS:969020979281924@1608044013881/A-Latitudinal-distribution-of-soils-used-in-this-study-where-smaller-bins-represent_Q320.jpg)
Content uploaded by Rebecca Dzombak
Author content
All content in this area was uploaded by Rebecca Dzombak on Dec 15, 2020
Content may be subject to copyright.
SoilSyst.2020,4,73;doi:10.3390/soilsystems4040073www.mdpi.com/journal/soilsystems
Article
WeatheringIntensityandPresenceofVegetationAre
KeyControlsonSoilPhosphorusConcentrations:
ImplicationsforPastandFuture
TerrestrialEcosystems
RebeccaM.DzombakandNathanD.Sheldon*
DepartmentofEarth&EnvironmentalSciences,UniversityofMichigan,AnnArbor,MI48109,USA;
rdzombak@umich.edu
*Correspondence:nsheldon@umich.edu
Received:2September2020;Accepted:13December2020;Published:15December2020
Abstract:Phosphorus(P)isanessentiallimitingnutrientinmarineandterrestrialecosystems.
UnderstandingthenaturalandanthropogenicinfluenceonPconcentrationinsoilsiscriticalfor
predictinghowitsdistributioninsoilsmayshiftasclimatechanges.WhileitisknownthatPis
sourcedfrombedrockweathering,relationshipsbetweenweathering,P,andothersoil‐forming
factorshavenotbeenquantifiedatcontinentalscales,limitingourabilitytopredictlarge‐scale
changesinPconcentrations.Additionally,whileweknowthatFeoxide‐associatedPisanimportant
Pphaseinterrestrialenvironments,therangeinandcontrolsonsoilFeconcentrationsandspecies
(e.g.,Feinoxides,labileFe)arepoorlyconstrained.Here,weexploretherelationshipsbetweensoil
PandFeconcentrations,soilorder,climate,andvegetationinover5000soils,andFespeciationin
ca.400soils.WeatheringintensityhasanuancedcontrolonPconcentrationsinsoils,withP
concentrationspeakingatintermediateweatheringintensities(ChemicalIndexofAlteration,
CIA~60).Thepresenceofvegetation(butnotplantfunctionaltypes)affectedsoils’abilityto
accumulateP.Contrarytoexpectations,PwasnotmorestronglyassociatedwithFeinoxidesthan
otherFephases.TheseresultsareusefulbothforpredictingchangesinpotentialPfluxesfromsoils
toriversunderclimatechangeandforreconstructingchangesinterrestrialnutrientlimitationsin
Earth’spast.Inparticular,soils’tendencytoaccumulatemorePwiththepresenceofvegetation
suggeststhatbiogeochemicalmodelsinvokingtheevolutionandspreadoflandplantsasadriver
forincreasedPfluxesinthegeologicalrecordmayneedtoberevisited.
Keywords:soilfertility;phosphoruscycling;weathering;ironspeciation;biogeochemistry
1.Introduction
Phosphorus(P)isanessential,oftenlimitingnutrientinecosystemsacrosstheglobe[1,2].Phas
beenstudiedinplants,soils,rivers,lakes,andoceans—bothmodernandancient—andisatthecenter
ofcriticalquestionsaboutEarth’spastandfuture.Today,concernsaroundPinterrestrialsettings
focusonsoilfertilityandcropyields,aswellasonagriculturalrunoffanditseffectoneutrophication
[3–5].Climate,vegetation,andweatheringanderosionratesareinextricablylinked,soinorderto
improvequantitativeconstraintsonpotentialPfluxesfromsoils,eachofthesefactorsmustbe
considered.Theredox‐sensitivemetaliron(Fe)isoftenassociatedwithbothPandorganic
matter/carbon(C)insoils(e.g.,[6–9]).TheP‐FeoxidepathwayiscriticalforterrestrialPtransport;
however,itiscurrentlylesswell‐understoodthanotheraspectsofterrestrialPcycling,suchasP‐C
associations.Additionally,whilemuchresearchlinkingP,weathering,andsoilagehasbeendone,
SoilSyst.2020,4,732of37
suchworkhastypicallydoneonrelativelysmallscales(e.g.,chronosequencesinHawaii)relativeto
theglobalscaleofthequestionsthatweaskhere.Therefore,constrainingclimate,vegetation,and
weatheringcontrolsonsoilPandFeiscriticalforpredictingchangestofluxesofPfromsoilstorivers
andbeyond,andtoimprovingourunderstandingofhowregionalsoilfertilitymayevolvein
responsetoachangingclimate.
Priortotheevolutionandexpansionoflandplants,changesinthefluxofPfromcontinentsto
oceansarethoughttohavebeenaprimarycontrolonmarineproductivity,andconsequently,onthe
concentrationofoxygenintheatmosphere[10,11].Itremainsdebatedwhetherlandplants’
colonizationofcontinentswouldhaveincreased[12]ordecreased[13]thatPflux.Thebetterwe
understandthemoderndistributionofandcontrolsonPinsoils,thebetterwecanmodelglobal
biogeochemicalchangesinEarth’spast.
1.1.P,Fe,andWeatheringandErosion
PisreleasedfrombedrockviatheweatheringofP‐bearingminerals(primarilyapatites,which
areCamineralsrelativelyresistanttoweathering;[4]).Weatheringdependsonallsoil‐forming
factors(climate,biology,topography,time,parentmaterial),butisprimarilycontrolledbyclimate
(e.g.,precipitation,temperature,seasonality)andtime(surfaceage,lengthofexposure;[14]).Organic
matteralsoservesasanimportantpoolofPinsoils([4,5]),andvegetation,fungi,andmicrobiallife
playanimportantroleinbothPandFeliberationandmobilizationthroughsymbiotic(mycorrhizal)
relationships[15,16].Thepresenceofplants,then,changeshowbothPandFearedistributedin
soils—butthelinksbetweenlandscape‐scalesoilchemistry,plantfunctionaltype,climate,and
weatheringintensityhavenotbeenquantified.
Pcanbedepletedfromasoilprofilewithinthousandstotensofthousandsofyears[4,17–20].In
terrestrialvegetation(biomass),P’sresidencetimeis~13years,andresidenceinsoilis~600years
withoutPreplenishmentbydust[4].SoilPcanalsobereplenishedthroughdustdeposition(e.g.,
[17,21,22]).Thenaturalprogressionofasoil,withaslowaccumulationofFeandAloxidesandaloss
ofmobilecationsandnutrients,meansthatPistypicallyverylowinoldand/orintenselyweathered
soils[19].Othersoilproperties,suchasclaycontent,arealsolinkedtoweatheringandvegetation,
andcouldaffecttheconcentrationsofbothPandFe.Claymineralsandclaygrain‐sizefractionscan
sorbconsiderableamountsofPandFe,andclaycontenttypicallyincreaseswithweatheringtime
and/orintensity[23,24].Weatheringintensityisexpectedtodecreasewithlatitudeduetocolder
temperatures,drierprecipitationregimes,andlessvegetationandshallowerrootingdepths[23,24].
Additionally,becausethepotentialvolumeoferodiblematerialgenerallyincreasesasweathering
intensityincreases(thoughvarieswithfactorssuchasslopeandupliftrate,e.g.,[25–29],weathering
islinkedtoPfluxesnotonlythroughdirectapatitedissolution,butalsothrougherosionrates,e.g.,
[30,31]).UnderstandinghowweatheringintensityisrelatedtosoilP(andFe)concentrationsis
importantforpredictinghowtheirerosionalfluxesandtransportmaychangeinresponsetoclimate
change(e.g.,[30,31]).
Afterweathering,dustdepositionisanimportantsecondarysourceofP,particularlyforold
soilswhosePhasbeendepletedthroughweathering,erosion,andbiologicaluse.Forexample,
chronosequencestudiesinHawaiihaveshownthatforoldersoils,dustreplenishmentofPiscritical
tomaintainingfertilityoncebedrocksourceshavebeendepleted[17,32,33].Dustisalsoacritical
sourceofPintheAmazonbasin,whichreceivessignificantdustfluxesfromAfrica[18,34–36],
compensatingforPlossindeeplyweatheredtropicalsoils.Dust‐sourcedPislesssignificantforareas
thatreceivelittledustdeposition,foryoungersoilsthatstillhavearegularinputofPfrombedrock
weathering,orforshortertimescalesthaneitherforsignificantdustaccumulationortimescalesof
soilslosingtheirbedrock‐sourcedP(i.e.,anthropogenictimescales).Arvinetal.[21]foundthatdust
depositionisanimportantPsourceinmontanesoils,despitetheirrelatively“fresh”bedrock.The
balancebetweenweatheringrateanddustaccumulationisalsorelevantfordetermininghow
importantthelatterisinsoilP[22].DustisalsocommonlyasourceofFeoxides,increasingsoil
fertility[35,37,38].
SoilSyst.2020,4,733of37
1.2.PandFeKineticsinSoils
PandFearechemicallyassociatedinsoilsinpartbecauseofP’saffinitytosorbtoFe
(oxyhydr)oxides(e.g.,[4,9,39–42]).Orthophosphate,thecommonionformoffreePinsoils,isreactive
towardsthoseoxidesandbecomesimmobile(notreadilybioavailable)whentheyassociateinto
neoformedmineralsoraresorbedontoexistingmineralsurfaces.Fe‐boundPmaybeacriticalmode
ofPtransportamongterrestrialecosystemsandfromterrestrial/aquatictomarineenvironments,as
oxyhydroxide‐boundPisoneofthemajorbioavailable/exchangeableformsofPinfluvialsystems
[5].
WhileweknowthatFeoxide‐adsorptionofPiskineticallyfavorableinsoilsandother
environmentswherePisbeingexchanged(e.g.,[43–45]),ourunderstandingoffirst‐ordercontrols
onFeconcentrationandspeciation(i.e.,Fe3+inoxidesversuslabileFe2+)insoilscouldbeimproved.
ExaminingafullsuiteofextractableFe(ratherthanonlydithionite‐citrateand/orammoniumoxalate)
allowsamorenuancedpictureofFephasesinsoils,whichcanultimatelybelinkedtovariableP
mobilityindifferentsoilredoxconditions[43,46,47].Itispossiblethatregionalclimatefactors
(precipitation,temperature)exertcontrolonFespeciation(i.e.,reducedvs.oxidized)insoilsviasoil
moisture.Controlsonsoilmoisturearedebatedandlikelyvaryfromprofiletoprofile[48–53],and
pH(relatedtosoilporespacesaturation;oftenmicrobially‐mediated)isaprimarycontrolon
reduction/oxidationofFephasesinsoils(e.g.,[54]).Additionally,soilporewaterscanserveasan
intermediatestepbetweenrecalcitranceandfluvialsuspensionfordissolvedconstituents(e.g.,
[54,55]).ExploringtheserelationshipsiskeyforunderstandingthelikelihoodofPmobilizationunder
differentmoisturesituations.
1.3.PandFeintheGeologicRecord
Geologistsareinterestedinconstrainingmanyoftheprocessesdescribedaboveinancientsoils
andecosystems.WeatheredP,transportedtotheoceansviafluvialnetworksandcontinental
drainage,istypicallyinvokedasafirst‐ordercontrolonmarineprimaryproductivity,whichis
associatedwithCsedimentation,atmosphericCO2drawdown,andoxygenproduction(e.g.,
[10,11,56]).SimilartoP,Fecanstimulatemarineproductivityonshorttimescales(e.g.,[57–60]),but
itisofadditionalinterestbecauseinthepast,itservedasaPsorptionsiteinmarinesystems,
effectivelysequesteringPsothatitisnotbioavailable(theso‐called“Fe‐Ptrap”)andlimitingmarine
productivity[10].
Withtheseprocessesinmind,geologistsareinterestedincontinent‐to‐oceanfluxesandmobility
ofFeandPthroughtime.However,thecontinent‐to‐oceantransportofP,althoughcentraltomany
biogeochemicalmodels,isusuallyprescribedandnotbasedonactualchangesinfossilsoil(paleosol)
Pvaluesorweatheringintensitydata.Rather,ithasbeengeneralizedbasedoncrustalPvaluesand
fluxescontrolledbylarge‐magnitudechanges,suchasglobalglaciationsorlimitedmodern
observations(e.g.,[61]).ThevalueinprovidingrobustconstraintsontherangeanddistributionofP
insoils,aswellasclimaticandenvironmentalcontrolsonitsmobilityandbioavailability,willvastly
improvetheparametersforthesemodels.
Inadditiontobiogeochemicalmodels,geologistsareinterestedinhowtheevolutionof
terrestrialvegetationandmycorrhizae(fungalrootsymbionts)affectedPbioavailability,terrestrial
mobility,andfluxestotheoceans.Researchershavearguedforbothanincrease[12,56,62,63]and
decrease[13,15]inPfluxestotheoceansfollowingthespreadoflandplants.Constraininghow
modernvegetationrelatestosoilPmobilityanddistributionswillhelptoanswerthisquestion.
Finally,thegeochemicalcompositionoffossilsoilsisusedtoreconstructarangeofclimaticand
environmentalchanges;however,thesetoolsareprimarilybasedonsmalldatasets.Thiswork
providesrangesofreasonablevaluesforsoilchemistriesunderknownenvironmentaland
atmosphericconditions,providingcriticalbackgroundinformationtoimproveourpaleoclimateand
paleoenvironmentreconstructions.
SoilSyst.2020,4,734of37
1.4.ThisWork
Inthecontextoftheseconcerns,severalbig‐picturequestionsremain.Despitemuchresearchon
Pinmodernsoils,ecosystem(vegetation,anthropogenicimpact)andclimate(precipitation,
weathering)controlsonitsconcentrationhavenotbeenquantifiedonlandscapetocontinentalscales
forgeneralizableconclusions.Additionally,althoughPisknowntohaveastrongaffinitytowardsFe
(oxyhydr)oxidesinsoils,Fespeciesandsoilredoxconditionshavenotbeenconsideredaspotential
first‐ordercontrolsonlandscape‐scalePbioavailability.Questionsabouthowclimate(precipitation,
temperature),soilfactors(drainage,soilmoisture,claycontent),andoverallweatheringintensity
affectsoilFe‐Passociationsneedstobeinterrogated,toseeifFespeciesultimatelyservesasa
mediatorforPbioavailability.Moreover,oncethedistributionofandcontrolsonsoilPhavebeen
quantified,howdothosenewconstraintsaffectorconstrainourunderstandingofbiogeochemicalP
modelsinEarth’spast?
Weaddressthesequestions,alongwithhypothesesaboutPinthegeologicrecord,usingalarge
geochemicaldatabaseofmodernsoils(n>5000).Theresultscaninformmodelsofbothpastand
futureterrestrialbiogeochemicalcycles.
2.MaterialsandMethods
2.1.Sampling
Soilsampleswerecollectedinthefieldbytheauthors,andreceivedfromtheUSDAKSSLsoil
repository,andfromcontributors(totaln=404).Amajority(n=247)ofthesesamplesareBhorizons,
andtherefore,reflectlonger‐term,stablepoolsofP(e.g.,PsorbedtoclaysorFe(oxyhydr)oxides,
whichwetestforhere,orrecalcitrantorganicmatter)andFe.ThesesoilsincludeEntisols,Inceptisols,
Alfisols,Ultisols,Histosols,Mollisols,andAridisolsandprovidecoveragefromlow(Hawaii)tohigh
(Iceland)northernlatitudes(SupplementalFigureS1).Soilscollectedinthefieldorfromrepositories
werescreenedforanthropogenicinfluence(i.e.,ondevelopedordisturbedland),with
anthropogenically‐alteredsoilsnotedassuchandbinnedseparatelyinourtreatmentsofthedata
here.Tocomplementthesesoils,weusedaUnitedStatesGeologicalSurvey(USGS)soilgeochemistry
interactivereport[64]andcollatedgeochemicaldatafortheTop5cm,Ahorizons,andChorizonsof
allthesoilsincludedinthatwork(n=4857).SeeSupplementalTableS1forsummarydescriptionsof
thesetwodatasets,andSupplementalTableS2forcompletesampleinformationforthenewdataset
presentedhere.GeochemicaldatafortheUSGSdataset[64]areavailableat
https://doi.org/10.3133/sir20175118.
Combiningthesesoildatabasesyieldsexcellentspatial,climatological,profiledepth,drainage,
vegetation(asplantfunctionaltype/generalizedbiome),andsoilordercoverage(Supplemental
FigureS1).WhilethesamplesaremostlyfromtheUnitedStates,variationinsoilformingfactors
acrossthecontinentreflectmostofthepotentialglobalvariation.Thetropicsaresomewhatunder‐
sampled,butdatafromHawaiiarebroadlyreflectiveofthoselatitudes.Whiletherearespecies‐
specificrelationshipsbetweenplantsandsoilgeochemistry,generalizedbiomeisacceptableforthe
continentalscaleofquestionsexaminedinthiswork,aswellasthecomparativenatureofour
inquiries(e.g.,[65–68]).Differencesbetweenplantfunctionaltypestypicallyswampdifferences
withinoneplantfunctionaltype.Additionally,becausebiomesonsub‐continentalscalesaretypical
forglobalbiogeochemicalmodels,andbecausemorespecificdataarenotavailableforthelatitudinal
scaleofthiswork,generalizingbybiomescale/plantfunctionaltypeisappropriate.
2.2.GeochemicalAnalyses
AllanalyseswereperformedondrysoilswithintheSheldonlab,asweresamplesduringthe
USGSanalyses.SoilsweredriedperUSDAAPHISregulationsuponimportandgroundto<70μm
forhomogeneityduringanalyses;sampleswerestoreddryinoxic(ambientroom)conditions(during
theexperimentaldesign,testsofpre‐andpost‐drying,withdryingtemperaturesvaryingbetween
25,50,and100C,showednosignificantgeochemicaldifferences).Bulkelementalgeochemistrywas
SoilSyst.2020,4,735of37
performedatALSLaboratoriesinVancouver,BC,Canada,usingafour‐aciddigestionand
InductivelyCoupledPlasmaMassSpectrometer(ICP‐MS)analysis.Externalprecisionwas0.01%for
allmajorelementsand10ppmforP2O5,analyzedatALS.ForsamplesfromtheUSGSdataset,1σ
sampleerrorswere488ppmand1.39wt%forP2O5andtotalFe,respectively)[64].The1σsample
errorsfortotalCandorganicCwere4.6%(n.d.forinorganicC[64]).
ToanalyzeFespeciationinsoils,weusedafour‐stepsequentialFeextractionfollowingPoulton
andCanfield[69]andRaiswelletal.[70].Thisextractionprotocolseparatesoperationally‐defined
poolsasascorbicacid(labileFe;Feasc),dithionite‐(crystallineFeoxyhydroxides;Fedith),ammonium
oxalate‐ (highlycrystallineFeoxyhydroxides;Feox),andsodiumacetate‐extractableFe(Fein
carbonates,analyzedseparatelyfromthesequentialextraction;Feacet).Initially,weincludeda
chromium‐reduciblesulfur(CRS;[71])extractionseparatelytoanalyzeforFesulfides(e.g.,pyrite),
butthoseconcentrationswerebelowdetectionlimitforallnon‐wetlandsoils.CRSwaslimitedto
wetlandsoilsaftertheinitialexploratoryextractions,andrawvalueswerenormalizedtoaveragesoil
orderdensitiestoaccountforthelargedifferenceindensitybetweenHistosolsanddenser,more
mineral‐richsoilorders[72].Samplesupernatantswerecentrifugedanddiluted50×,andFewas
measuredviaICP‐OpticalEmissionSpectroscopyandICP‐MSattheUniversityofMichigan.
Aninternalstandardofalocalsoil,driedatthreedifferenttemperatures,wasincludedinall
runsforinter‐runconsistencychecksandtodeterminereproducibility.Theabsolutestandard
deviationforasample’sreproducibilitydependsontheconcentrationofeachFespecies,sothe
relativepercenterror((σ/μ)×100)isusedinstead.RepeatedanalysesshowthatFeascisreliablewithin
3.7%ofthemean,Fedith,within1.7%ofthemean,andFeox,within7.0%ofthemean.Reproducibility
ontheFeacetsampleswaslarger,at15%ofthemean,becausethemeansinmanyofthesampleswe
analyzedwerenearmachinedetectionlimits(~1000ppm),whereadifferenceofseveralhundred
ppmyieldsfarlargerstandarddeviationsthanspecieswithhigherconcentrations.TheFeascpoolis
consideredtobealowconstraintbecausealargerproportionoflabileFemaybeinasoil’sporewater
(ratherthansolids).Becausehereweareprimarilyinterestedinrelativedifferencesbetweenoxide
andlabileFespecies,andbecausespecies’concentrationstendtodifferbyordersofmagnitude,we
havedeemedthisapproachacceptable.Forfurtherreadingonnuancesinsamplepreparationand
extractionprotocols,seetworecentreviewsbyRaiswelletal.[73]andAlgeoandLiu[74].
FespeciationdatawerefilteredforoutliersbyremovinganysampleswhereFespecweight
percent>totalFeweightpercent(i.e.,theconcentrationofameasuredFepoolwashigherthanthe
measuredtotalFe).Thesediscrepancies(n=107,95,139,106,forFeasc,Feacet,Fedith,andFeox,
respectively)likelyarosefromthe50×‐dilutionbeinginsufficient,withhighFeconcentrations
overwhelmingtheICP‐OES,leadingtoerroneousmeasurements.FutureworkusingFespeciation
onsoilsshouldconsiderusingstrongerdilutionfactorsforhigh‐Fesoilstocircumventinstrument
limitations.
2.3.OtherDataCollated
GeochemicaldatafortheTop5cm,Ahorizons,andChorizonsfrom4857soilsaroundtheU.S.
areavailableonaninteractiveportal[64].Weathering‐relevantcations(Al,Ca,Na,K)andthe
nutrientsofinterest(P,Fe)alongwithlatitudeandtotalclaycontent(clay:AandChorizonsonly)
werecompiled.FortheUSGSdataset,claycontentwasdeterminedbyX‐Raydiffraction(for10and
14Åclaysonly).Organic,inorganic,andtotalcarbondatawereavailableforamajorityofsoilsinA
andChorizonsintheUSGSdataset(notallhadinorganiccarbon).Itshouldbenotedthatadrysoil
samplewillhaveahigherrelativeweightpercentorganiccarbonduetowaterlossduringdrying.
Soilordersanddrainagecapabilityarenotgivenforsoilsinthisdatabase,butbecausetheyspanthe
conterminousU.S.,itisareasonableassumptionthattheyincludeavarietyofsoilorders.Clay
contentasgrain‐sizefraction,parentmaterial,andsoilmoistureregimewerealsogatheredforthe
newly‐collecteddatasetfromUSDASoilSeriesreportswhenavailable(SupplementalTableS2).
ClimatedataweregatheredfromSoilSeriespages;whenthesewerenotavailable,thePRISM
30‐yearaverages(1981–2010)wereused.IfneitherSoilSeriesnorPRISMclimatedatawereavailable
SoilSyst.2020,4,736of37
(i.e.,fornon‐USsamples),country‐specificgovernmentalmeteorologicaldataorjournalarticleswere
used(seeSupplementalTableS2fordetails).
TheChemicalIndexofAlteration(CIA)[75]wascalculatedtoassessthedegreeofweathering
foreachsoil,whichreflectsallofthesoil‐formingfactorstosomedegree,butpredominantlysoilage
andclimate[14].TheCIAisthoughttoreflecttheweatheringoffeldsparstoformclaymineralswhere
highvaluesrepresentmoreintenseweathering.Molaroxideratiosareused(Equation(1)).TheCIA
ofanindividualsoilmaynotcorresponddirectlywiththeCIAvaluesofsedimentsinrivers(e.g.,
[76]).However,becauseoursoils’averageCIAvalues(59±15,42±32,and57±21forA,B,andC
horizons,respectively)arewithin1σoftheNorthAmericanrivers’averagefromthatstudy(n=7;
meanCIA=66±8.5,range53–73),thoughwithlargerranges(0.5to99forallthreehorizons),we
proceedwithitsusehere.
CIA=(Al2O3/(Al2O3+CaO+NaO+K2O))×100(1)
2.4.StatisticalAnalyses
ToexaminehowthetotalFeandPcorrelatewithCIA(weatheringintensity),arunningaverage
witha10‐CIA‐unitwindowwasused.Linearleast‐squaresregressionswereusedtotestfor
significantlinearcorrelationswhereappropriate,basedonthehypothesesthatwerebeingtested.
DuetothelargesizeoftheUSGSdataset,p‐valuesareextremelylow(<10−16;essentially0),meaning
itishighlylikelythatthetwovariablesarenotrandomlycorrelated.Strengthofcorrelationshould
beprimarilyconsideredbeforetakinganyrelationshipstobepredictive,butforsimplicity’ssake,
coefficientsofdetermination(R2values)arereported.PrincipleComponentsAnalysis(PCA)was
performedforUSGSdatabasesoilsbyhorizon(Top5cm,A,andC).Theseanalysesareusedtotest
theexpectedrelationships(e.g.,thatFeandPwillbepositivelyandlinearlyrelated),ratherthanto
buildpredictivemodels.AllofthestatisticalanalyseswereperformedinMatlab.
3.Results
CompletesampleinformationandallgeochemicalresultsarefoundinSupplementalTablesS2–
S5andareavailableintheonlineversionofthispaper.ForthelargerUSGSdataset[64],wehave
geochemical,drainage,andvegetation(generalizedplantfunctionaltype)data,butnosoilorderor
Fespeciationinformation.Resultsforsoilorder,drainage,parentmaterial,andFespeciationanalyses
arebasedonlyonthenewlycollecteddataset(n=404).
3.1.FeandP:Concentrations,SoilOrder,andVegetation
3.1.1.P
TheaveragetotalPconcentration(referredtosimplyas“P”hereonwards)inthecontinental
crustisca.870ppmP,andthereisrelativelylittlevariationbetweenbedrocklithologies[77].The
maximumsoilPconcentrationinthiscompilationwas>20,000ppm(>2wt%;Chorizonofan
anthropogenically‐modifiedsoil),anorderofmagnitudegreaterthanthecrustalaverage.MeanP
concentrationsforeachhorizonstudiedweredepletedrelativetothecrustalaverage:Top5cm,A,
andChorizonsmeanswere660,632,and508ppmP,respectively(USGSdata),andBhorizonsfrom
thenewdatasethadameanof937ppmP.Pconcentrationsvaryspatiallythroughoutcontinental
U.S.(SupplementalFigureS2),withhotspotsinavarietyofdifferentgeologic,climatic,andbiologic
provinces.Pconcentrationsvariedamongsoilorders(Figure1),withthehighestconcentrationsin
Inceptisols(Figure1A),arelativelyweakly‐developedsoiltype.
SoilSyst.2020,4,737of37
Figure1.PandFeconcentrationsinmodernsoils’Bhorizons,binnedbysoilorder.Redlinesare
medianvalues,bluebinsare25thand75thquartile,andredcrossesarebeyondthe75thquartile;the
horizontaldashedgraylineisthecrustalaverage(870ppmPand3.5%Fe)[77].(A)Pconcentrations,
showingthatyoungersoils(Entisols,Inceptisols)tendtohavehigherPthanoldersoils(Alfisols,
Ultisols).OxisolshavehighrangesofP,butamuchsmallerbinsizethanotherorders(n=8).Andisols
havehighinheritedPfromvolcanicparentmaterials.(B)Feconcentrations,showingconsistentFe
concentrationsbetweenca.1and10wt%withtheexceptionofOxisols,whicharedefinedbytheir
highratesofFeoxideaccumulation.Binsizes:Entisols(44),Inceptisols(75),Alfisols(42),Ultisols(6),
Oxisols(8),Aridisols(76),Mollisols(60),Andisols(15),Spodosols(9),Gelisols(30),Unknown(39).
ThevegetationgroupsareBarren(lackingvegetation),Altered(i.e.,affectedbyhumanactivity),
Forest(nodifferentiationbetweendeciduousandconiferous),Herbaceous/Grassland,
Developed/Cultivated(i.e.,occupiedormanicured),Shrubland,andOther(mostlyincludingearly‐
successionplantsormicrobialearths).ForUSGSsamples,protocolsdictatedthatroads,buildings,
andindustrialsitesbeavoided[64].Amongvegetationtypes,thereislittlevariabilitybetweenP
concentrationsandvegetationtypes(Figure2);whileBarrenlandscapes’Bhorizonsshowhigher
meanP,thisislikelyduetosmallbinsize(n=5;Figure2C).Overall,variabilityinconcentrationsin
differentvegetationtypesandsoilorderswasminimal,withmostsoilsbeingdepletedinPrelative
tothecrustalaverage.Vegetatedsoils(bothnaturalandanthropogenically‐altered)hadhigher
rangesofPthanunvegetated(Barren).Therewasaweakpositivecorrelationwithmeanannual
precipitation(SupplementalFigureS3A),buttherangeofPvaluesatagivenprecipitationamountis
SoilSyst.2020,4,738of37
typicallytoolarge(~1500ppm)tobeofpredictiveuse.Thistrendcouldalsobeexplainedbyrelated
factors,suchasvegetationcoverage(coverversusbareground)andweatheringintensity,whichare
associatedwithprecipitation.
Figure2.
P
concentrationsinallhorizons,binnedbyvegetationtype.Dashedgraylineinallisthe
crustalaverageP(870ppm)[77].Arrowsindicateoff‐plotvalues.(A)
P
intheTop5cm,whereBarren
(novegetation)andAlteredsoilshavethelowestrangeofPconcentrations.(B)
P
inAhorizons,with
similardistributionstoTop5cm.(C)
P
inBhorizons,withmuchsmallerbinsizesthanotherhorizons.
(D)P
inChorizons.Binsizesfor(A,B,D):Barren(68),Altered(209),Forest(1252),Herbaceous(815),
Developed(1568),Shrubland(945).Binsizesfor(C):Forest(43),Grassland/Herbaceous(19),
Shrubland(50),Barren(5),Cultivated(59),Unknown(74).
3.1.2.Fe
ThecrustalcompositionoftotalFe(Fe
tot
)ismorevariablethanthatofPbecausewhilePis
sourcedprimarilybyapatite‐groupminerals,Fe
tot
canbepresentinawiderangeofmineralsand
lithologies.ThePhanerozoicuppercontinentalcrustisestimatedtohave3.5wt%Fe[77].Fe
tot
averagesinthesoilsstudiedherewere2.1wt%forTop5cm,1.6wt%forAhorizon,2.6wt%forC
horizon,and4wt%forourdataset(primarilyBhorizons).Density‐normalizedFe
tot
variedslightly
bysoilorder(SupplementalTableS3),showingtheexpectedtrendofmodestlossshiftingtomodest
accumulationduringtheAlfisol‐Ultisoltransition(Figure1B),butotherwisewasrelatively
consistent.OxisolshadthehighestvaluesandrangeinFe,withsomevariabilityamongtheothersoil
ordersbutgenerallywithinconsistentranges(Figure1B).
SoilSyst.2020,4,739of37
Fe
tot
isgenerallyconsistentbothwithinagivenhorizonandbetweendifferentvegetationcover
types(Figure3).Whilemostsoilordersandmosthorizonshad<3wt%Fe
tot
,valuesrangedupto
>15%(Figure3;SupplementalFigureS2).TheonlynotablevariabilityinFeconcentrationsamong
vegetationisinCultivatedsoilBhorizons(Figure4C),whicharedepleted,and‘Other’soilBhorizons
thataresubstantiallyenrichedrelativetotheothervegetationcovergroups.‘Other’vegetation
containsmostlybasalt‐parentedsoilsfromalimitedgeographicalarea(primarilyIceland),sothis
resultislikelyanartifactofoursampling.TherewasnostrongcorrelationbetweenFe
tot
inasoiland
meanannualprecipitation(R
2
:0.2;SupplementalFigureS3B).
Figure3.Feconcentrationsinallhorizons,binnedbyvegetationtype.Dashedgraylineinallisthe
crustalaverage(3.5%)[77].Arrowsindicateoff‐plotvalues.(A)FeintheTop5cm,whereBarrenand
AlteredsoilshavethelowestmaximumFeconcentrations.(B)FeinAhorizons,withsimilartrends
toTop5cm.(C)FeinBhorizons.TherewerenodataforBarrenBhorizons’Fe.(D)FeinChorizons.
Binsizesfor(A,B,D):Barren(68),Altered(209),Forest(1252),Herbaceous(815),Developed(1568),
Shrubland(945).Binsizesfor(C):Forest(43),Grassland/Herbaceous(19),Shrubland(50),Barren(0),
Cultivated(59),Unknown(74).
3.2.FeandP:Latitude,Weathering,SoilOrder,andClayContent
Latitude,weatheringintensity,andclaycontentshouldeachbeassociatedwitheachotheras
wellaswithFe(e.g.,[24]).Someofthetrendsdescribedherecouldbeinfluencedbyalatitudinal
samplingbias,withmostofthesamplescomingfromthemid‐latitudes(SupplementalFigureS1;
SoilSyst.2020,4,7310of37
Figure4A).Inthesoilsanalyzedhere,maximumweathering(asmeasuredbyCIA)decreasesas
latitudeincreases(Figure4B),whileclaycontent(aweatheringproduct)doesnotshowastrongtrend
withlatitude(Figure4C),withtheexceptionofaweakmid‐latitude(35–40°N)peakinclaycontent
insomeChorizons.Weatheringtrendsbetweensoilordersbehaveasexpected,withCIAvalues
increasingfromEntisolstoUltisols/Oxisols(Figure5).Claycontentandweatheringshowawell‐
behavedandexpectedpatternofincreasethatmatchesthetheoreticalunderstandingofthatmetric
[24,75,78,79],withBhorizonsspecificallyshowingthehighestclaycontentuntilthehighestCIA
valuesarereached(Figure4D,greenpoints).Claycontentandgeochemicaldatawerenotavailable
fortheTop5cmsubsetoftheUSGSdataset,sothathorizonisexcludedfromthosecomparisons.
Figure4.(A)Latitudinaldistributionofsoilsusedinthisstudy,wheresmallerbinsrepresentthe
USGSdataset[64]andthewiderbinsareoursoils.Therangeoflatitudescoveredhereis7°to73°N,
butmostofthesamplesfallbetween20and50°N.(B)LatitudinaltrendsinChemicalIndexof
Alteration(CIA).(C)Latitudinaltrendsinclaycontent.(D)CorrelationbetweenCIAandclaycontent
(R
2
:0.05,<0.01,and0.19forA,B,andChorizons,respectively).Overall,theseresultssupportcommon
assumptionsabouttherelationshipsbetweenlatitudeandweathering.
SoilSyst.2020,4,7311of37
Figure5.ChemicalIndexofAlteration(CIA)binnedbysoilorder.CIAincreasesthroughtheEntisol‐
Ultisolsoildevelopmentprogression,asexpected,andOxisolshavethehighestCIAwithamedian
near100.Aridisols’medianCIAfallsbetweenEntisolsandInceptisols.Binsizes:Entisols(44),
Inceptisols(75),Alfisols(42),Ultisols(6),Oxisols(8),Aridisols(76),Mollisols(60),Andisols(15),
Spodosols(9),Gelisols(30),Unknown(39).
Fe
tot
contentincreasesmoderatelywithhigherlatitudes(Figure6A),higherweathering(Figure
6C),andclaycontent(Figure6E).Pinsoilsbehaveslesslinearlywithrespecttothesevariables:P
increasesmoderatelywithlatitude(Figure6B),butratherthandecreasingwithweathering,asmight
beexpectedbasedonsoilage‐Prelationships,averagePconcentrationsinallhorizonsvarylittle,
thoughthereisageneralincreaseinrangeatmoderateweatheringintensities(CIA~40–70)(Figure
6D).Pconcentrationsareweaklynegativelycorrelatedwithclaycontent(Figure6F).Fe
tot
andP
showedpositivecorrelationswithoneanotherinallhorizons,asexpected(SupplementalFigureS4).
SoilSyst.2020,4,7312of37
Figure6.Trendsbetweenlatitude,weathering,andclaycontentandtotalFeandPconcentrations.
Solidlinesin(C)and(D)arerunningaveragesforTop5(yellow),A(red),B(blue),andC(black)
horizons.Solidlinesin(E)and(F)arelinearleast‐squaresregressionsforA(red),B(blue)andC
(black)horizons.(A)Feinsoilsbylatitude.(B)Pinsoilsbylatitude.(C)Fe
tot
andweatheringintensity.
Solidlinesaretherunningaverages(10‐CIA‐unitwindow)foreachhorizon;colorscorrespondto
datacolors.TheBhorizonaverage(blueline)ishighestatveryhighCIAsdueinparttolowersample
densityatveryhighCIAvalues,especiallybetween98–100.(D)Pconcentrationsandweathering
intensity.(E)Feandclaycontent(R
2
is0.34and0.37forAandChorizons;0.01forBhorizons,much
smallerdataset;p=0.7forBhorizon,<<0.01forotherhorizons).Solidlinesarelinearleast‐squares
regressions;theBhorizon(blueline)isskewedduetohigh‐Fe,low‐claypointsoffplot.(F)Pandclay
content.Solidlinesarelinearleast‐squaresregressions(R
2
is0.11,0.08,and0.01forA,B,andC
horizons,respectively;p<<0.01forallhorizons).
SoilSyst.2020,4,7313of37
3.3.Weathering,ClayContent,andVegetation
Betweenvegetationtypes,thereislowvariabilityinclaycontent,withForestsshowingslightly
lowervaluesinBhorizonsthantheothergroups(SupplementalFigureS5).Forestshavethehighest
weatheringinnatural(unaltered/notdeveloped)soils,followedbyGrasslandsandShrublands
(Figure7).
Figure7.ChemicalIndexofAlteration(CIA)binnedbyvegetationtype,forhorizons(A–C)(bulk
chemistrydataforCIAwerenotavailablefortheTop5cmhorizon).(A)Alteredsoilshavethehighest
CIAvaluesoverall,andForestsoilshavethehighestCIAfornon‐anthropogenically‐affectedsoils,in
Ahorizons.(B)ForestshavethehighestCIAvaluesinBhorizons,followedbyGrasslandsand
SoilSyst.2020,4,7314of37
Shrublands(Barrensoils’smallbinsize,n=5,precludesitfromanalysishere).(C)Alteredsoilsalso
havethehighestCIAinChorizons,followedbyForests.
3.4.FeandP:DrainageandSoilMoisture
Moderatelypoorly‐drainedsoilshadhigherFeandPconcentrationsthantheotherdegreesof
drainage.PerudicmoistureregimeshadthehighestFeandPvalues(Figure8);however,these
samplesweredominatedbybasalt‐parentedsoilsinIceland,whichmayskewtheresults.Asidefrom
that,AquicandXericmoistureregimeshadhighFe,andUsticandUdicsoilshadhighP(Figure8).
SometrendsemergewithrelativelyhighFe/lowP(Aquic)andviceversa(Udic).
Figure8.VariabilityinFeconcentrations(A,C)and
P
concentrations(B,D)indifferentsoildrainage
andmoisturesregimes.DashedgraylinesrepresentcrustalaveragesforFeandPinrespectiveplots.
(A)Febinnedbydrainage.(B)
P
binnedbydrainage.(C)Febinnedbymoistureregime.(D)
P
binned
bymoistureregime.SomepatternsemergewhencomparingFeandPbetweensoilmoistureregimes,
e.g.,UdicsoilshavinglowFe/highPandAquicsoilshavinghighFe/lowP.Perudicsoilsare
dominatedbybasalt‐parentedsoils,influencingtheirFeandPconcentrationsandhighlightingthe
importanceofbedrockparentmaterial.NodataforPermafrostFeorPconcentrations.Binsizesfor
(A,B):Poorly‐drained(166),Moderatelypoorlydrained(25),Moderatelywell‐drained(16),Well‐
drained(185),Excessivelywell‐drained(12).Binsizesfor(C,D):Unknown(111),Aridic(73),Aquic
(15),Udic(82),Perudic(27),Ustic(54),Xeric(21),Permafrost(N.D.).
SoilSyst.2020,4,7315of37
3.5.FeSpeciation
3.5.1.FeSpeciationandP
Pyrite/Feinsulfides,normalizedtoaveragesoilorderdensity,wereabovedetectionbytitration
inonly22testednon‐wetlandsoilsoutof63fromaroundthecontinentalU.S.(SupplementalTable
S5),sothattestwasexcludedfromsubsequentanalyses.Density‐normalizedpyriteyieldsfromall
wetlandsedimentsanalyzed(n=15)weremeasurable/abovedetectionlimit(>0.1wt%),asopposed
tonon‐wetlandsoils(Alfisols,Aridisols,andMollisols),whichtypicallyfellbelowthislimitafter
beingdensity‐normalized.Inanon‐perennially‐waterloggedsoil(e.g.,notwetlands),itisreasonable
toassumethatFecontentsinsulfidesarenegligible.
Contrarytoexpectations(seeSection4.3below),Fe
dith
andFe
ox
poolsdidnotdisplaystrongeror
morerobustcorrelationswithP(Figure9)thantheotherFepools.Therewerenoclearcorrelations
betweenPandanyFespecies(forall,R
2
<0.01).SeeSupplementalTableS4forallFespeciesresults
andFe
3+
/Fe
2+
ratios.
Figure9.RelationshipsbetweenFespeciesandPconcentrations,allwithR
2
values<0.1.Arrows
indicateoff‐plotvalues.(A)NostrongcorrelationbetweenFe
asc
andP(p=10
−6
).(B)Nosignificant
correlationbetweenFe
acet
andP(p=0.2).Notethedifferentx‐axisscale.(C)Nostrongcorrelation
betweenFe
dith
andP(p=0.003),whereapositivecorrelationhadbeenexpectedduetoP’saffinityto
sorbtoFe(oxyhydr)oxides.(D)NostrongcorrelationispresentbetweenFe
ox
andP(p=10
−9
),again
whereastrongpositivecorrelationmighthavebeenexpected.
SoilSyst.2020,4,7316of37
3.5.2.FeSpeciation,Precipitation,SoilMoisture,andDrainage
Contrarytoexpectations(seeSection4.2.2below),Fespeciationpoolsshowednostrong
predictiverelationships(allR
2
<0.2)withmeanannualprecipitation(Figure10),buttherewere
differencesbetweensoilmoistureregimes.PermafrostsoilsfromtheNorthSlopesofAlaska,highin
organicmatterandFe,weredominatedbylabileFe(Fe
asc
).AquicandUdicsoilshadhighFein
carbonates(Fe
acet
),andPerudicsoils(primarilyfromIceland,withbasaltparentmaterial)hadhigh
Fe
ox
(Figure11).MeanFe
dith
wasconsistentbetweensoilmoistureregimesbuthasanextremelyhigh
rangeinAridicsoils.
Figure10.Relationshipsbetweenmeanannualprecipitation(MAP;mmyr
−1
)andFespecies,colored
bysoildrainage.Arrowsindicateoff‐plotvalues.(A)NocorrelationbetweenMAPandFe
asc
(R
2
<0.1,
p=10
−9
).(B)NostrongcorrelationbetweenMAPandFe
acet
(R
2
<0.01;p=0.0005);therecouldbea
maximumMAPvaluebeyondwhichFe
acet
islesskineticallyfavorable.Notethedifferenty‐axisscale.
(C)NosignificantcorrelationbetweenMAPandFe
dith
ispresent(R
2
<0.01,p=0.7).(D)Modest
significantcorrelationbetweenMAPandFe
ox,
(R
2
=0.16,p=10
−17
).ForallFespecies,thereareno
strongpredictivetrendsbetweenFespeciesconcentrationandMAP,andnopatternsindrainage.
SoilSyst.2020,4,7317of37
Figure11.Fespeciesbinnedbysoilmoistureregime.(A)Udic,Perudic,andPermafrostsoilshavethe
highestmedianFe
asc
.(B)Fe
acet
isquitelowinmostsoils.Notethedifferenty‐axisscale.(C)Aridicsoils
havehighFe
dith
,butothermoistureregimesshowlittlevariability.(D)Littlevariabilitycanbeseenin
Fe
ox
acrosssoilmoistureregimes.Binsizes(slightvariabilitybetweenFespeciesisduetoextraction
yields):(A)Unknown(106),Aridic(73),Aquic(15),Udic(82),Perudic(26),Ustic(53),Xeric(21),
Permafrost(21).(B)Unknown(111),Aridic(73),Aquic(15),Udic(82),Perudic(27),Ustic(54),Xeric
(21),Permafrost(18).(C)Unknown(111),Aridic(61),Aquic(14),Udic(78),Perudic(26),Ustic(48),
Xeric(15),Permafrost(21).(D)Unknown(106),Aridic(73),Aquic(15),Udic(82),Perudic(27),Ustic
(54),Xeric(21),Permafrost(21).
Poorly‐drainedsoilshadahighrangeofFe
asc
andFe
dith
,whilewell‐drainedsoilshadhigh
amountsofFe
acet
(Figure12A).TheotherFespeciesanddrainageswereconsistent.Theroleofslope
wasconsidered,butlocal,small‐scaletopographicreliefwasnotreadilydeterminableformost
samples.
SoilSyst.2020,4,7318of37
Figure12.Fespeciesbinnedbysoildrainage.(A)Excessivelywell‐drainedsoilshaveverylowFe
asc
.
(B)Notethedifferenty‐axisscale.Fe
acet
ishighestinwell‐drainedsoils,althoughpoorly‐drainedsoils
showanunexpectedlyhighrangeaswell.(C)ExcessivesoilsshowthelowestmaximumFe
dith
,with
theotherdrainagesshowingsimilarmediansandranges.(D)Fe
ox
hassimilarpatternstoFe
dith
,with
threeveryhighvaluesformoderatedrainages.Binsizes:(A)Poorly‐drained(104),Moderatelypoorly
drained(16),Moderatelywell‐drained(11),Well‐drained(140),Excessivelywell‐drained(3).(B)
Poorly‐drained(109),Moderatelypoorlydrained(16),Moderatelywell‐drained(11),Well‐drained
(142),Excessivelywell‐drained(3).(C)Poorly‐drained(101),Moderatelypoorlydrained(15),
Moderatelywell‐drained(9),Well‐drained(123),Excessivelywell‐drained(3).(D)Poorly‐drained
(109),Moderatelypoorlydrained(16),Moderatelywell‐drained(11),Well‐drained(142),Excessively
well‐drained(3).
3.5.3.FeSpeciation,Vegetation,andSoilOrder
ForestsoilshadhigherFe
acet
(Figure13B)thantheothervegetationgroups,butFespecies
concentrationswereconsistentotherwise.OxisolshadhigherFe
dith
andFe
ox
thanothersoilorders
(Figure14C),butFespecieshadlittlevariabilitybetweensoilordersotherwise.
SoilSyst.2020,4,7319of37
Figure13.Fespeciesbinnedbyvegetationtype.(A)ForestsandGrasslandshavehigherFe
asc
than
othervegetations(exceptforUnknown).(B)ForestshavethehighestFe
acet
values(medianandrange).
(C)NopatternseeninFe
dith
andvegetation,exceptthatBarrensoilshavelowFe
dith
.(D)Nopatterns
inFe
ox
andvegetation.
SoilSyst.2020,4,7320of37
Figure14.BoxplotsofFespeciesbinnedbysoilorders,wherecenteredredcirclesarethemedians,
thedarkblueboxesarethe25thinnerquartiles,thewhiskersaretheouter25thquartiles,andthered
circlesarethestatisticaloutliers(outside3σ).(A)NodifferencesinFe
asc
betweensoilorders.(B)
UltisolshavethehighestFe
acet
,althoughAlfisolshaveahighrangeofFe
acet
aswell.(C)Oxisolshave
thehighestFe
dith
,asexpected.(D)OxisolshavethehighestFe
ox
,asexpected.
3.6.OrganicCarbon
Organiccarbon(C
org
)inAhorizonsrangedfrom0to61wt%,andinChorizons,rangedfrom0
to43wt%[64].TheaverageC
org
:PratioforAhorizonswas17.9:1,andforChorizons,18.3:1.
RelationshipsbetweenC
org
andCIA,P,Fe,andclaycontentwereweak(SupplementalFiguresS6and
S7).TherewerenostrongcorrelationsbetweeninorganicCandFe
tot
,orbetweenC
tot
andFe
tot
(SupplementalFigureS7).
3.7.ParentMaterial
ThereisslightvariabilitybetweenFe
tot
andPconcentrationsandparentmaterial,withigneous
bedrockandash/volcanicshavingthehighestmedianvaluesforFe,withhighPvaluesaswell
(SupplementalFigureS8).WhilelimestoneparentmaterialsshowhighFe,thereareonlytwosamples
inthisgroupandtheyarenotincludedinthisdiscussion.Theparentmaterialdoesnotappeartobe
apredictivecontrolineitherFe
tot
orP(SupplementalFigureS8).However,thiscouldbeduetothe
mixedprovenanceofmanymodernsoils’parentmaterial(i.e.,alluvium/colluvium,glacialdeposits,
etc.asopposedtocompositionallyhomogeneousbedrock).Amoretargetedexplorationinto
bedrock‐parentedsoils,Fe
tot
,andPcouldwellshowmorewell‐behavedrelationships.
SoilSyst.2020,4,7321of37
3.8.PrincipleComponentsAnalysis
Resultsfromtheprincipalcomponentsanalyses(PCA)fortheUSGSdataareshownin
SupplementalFigureS10.Principlecomponents1and2explain~60%ofthevarianceatamaximum.
However,alleigenvaluesareverylow(<0.6)anddataareessentiallyclusteredaroundtheorigin,so
whilethereareassociations,theyareveryweakandshouldbeinterpretedconservatively.Latitude,
P,Fe,andAlwereassociatedwithPC2intheTop5cmandAhorizons,andvegetationandCIAwere
associatedwithPC1.GroupingschangedintheChorizons,withFeandAlassociatedwithPC2,but
PandlatitudeassociatedwithCorgalongPC1.Inallhorizons,CIAandvegetationvectorsareopposite
toeachother,andwereorthogonaltosubparalleltomostothervectors.
4.Discussion
Inthissection,weexploreourresultswithaseriesofquestionsfocusedonconstrainingcontrols
onFeandtotalPinsoils,definingtheirrelationshipsoncontinentalscales,andmakinginferences
abouthowthoserelationshipsmayimpactterrestrialPfluxes.Theyarecenteredaroundkeysoil‐
formingfactors,suchasclimate,vegetation,andweatheringintensity,aswellassoilredox‐specific
factors(i.e.,precipitation,moistureanddrainage,andFespecies).Becauseamajorityofsamples
comefrombetween20and50°N,theinterpretationsmadeherearemostrobustforthoselatitudes.
4.1.HowdoLatitude,Weathering,andClayContentAssociatewithPandFeConcentrations?
Mostoftheexpectedrelationshipsbetweensoilorder,FetotandP,weathering,andlatitudewere
supported.Asexpected,weatheringgenerallydecreaseswithlatitude,droppingfromamaximum
CIAof>95at30–35°NtoaCIA~75at50°N(thenorthernlimitofthelargeUSGSdataset).Farther
north,someBhorizonsdeviatefromthatpattern,withCIAvalueshigherthanmightbeexpected
basedontheirclimaticregime[80,81].Interpretingthelatitudinaltrendshereshouldincludeacaveat
forlatitudinalsamplingbiasandlimitationsofthedatasetto,mostly,between20°and50°N.While
thePCAresults(SupplementalFigureS10)donotsupportarelationshipbetweenPandweathering,
weinterpretthisdiscrepancyasbeingduetothecomplexitieswithintheP‐weatheringrelationship
ratherthannegatingtheobservationsmadebetweenlatitude,weathering,andPconcentrations
becausethePCAeigenvaluesaresolow(most<±0.2),andessentially,areclusteredontheorigin,
whichindicatesanot‐predictivevalue.P’srelationshipwithweatheringiscomplexandduetoa
varietyoffactors(climate,time,slope/erosionrate,etc.),andbylookingonlyattheendproduct—
whichiswhatisleftinthegeologicrecordforanalysis—wemustinherentlyworkaroundthose
limitationsanddrawthemostrobustconclusionspossiblefromlimiteddataontheselargescales.
Fetotbehavesasexpected,accumulatinginBhorizonsassoilweatheringincreases,whileP
accumulationinallhorizonspeaksatmoderately‐weatheredsoils(CIA~60).Thelatterrelationship
isexpectedbecauseasoilthathasexperiencedmoderateweatheringandhasdevelopedsomewhat
(e.g.,Inceptisol,Alfisol)hashadsufficienttimetohavePmobilizedfromthebedrock/substrateand
bioticallycycled,butnotsolongorwithsuchintenseweatheringthatPisdepletedfromthesubstrate
andremovedviaerosionand/orlossofbiomass.Anolderand/ormoreintenselyweatheredsoil(e.g.,
Ultisol)willhaveasubstratemoredepletedofPandwillhavelostmoreP.IntermsofterrestrialP
transport,amoderately‐weatheredsoilisthemostlikelytohavealargepoolofpotentialPto
transport.Aninterestingnextstep,buildingoffthiswork,couldbetotesttheseexpectedcorrelations
bymappingsoilPwithsoilageandcollectingfluvialPloads.
Pconcentrationsare,onaverage,farbelowthecrustalaverageof870ppmP.Evenaccounting
fordensitydifferencesbetweenatypicalsoil(soil~1.62±0.2gcm−3;n=659)[72]andcontinental
crust(e.g.,granite;crust~2.7gcm−3),allofthesoilhorizonsaredepletedinPrelativetothecrustal
abundance.Thereissomevariabilitybetweensoilorders,butallbutasmallsubsetwerebelowthe
crustalabundance(i.e.,recentbasalt‐parentedsoils).Whenconsideringbulkdensity,thetotalFe
concentrationsshowtheopposite,withthesoilFeaverageof4.4%beinggreaterthanthecrustal
averageof3.5%,againwithsomevariabilitybetweensoilordersbutoverallgreaterthanthecrustal
average(SupplementalTableS3).Thishasimplicationsforbiogeochemicalmodeling(seeSection
SoilSyst.2020,4,7322of37
4.7.3).WhilePconcentrationscouldbeexpectedtogenerallydecreasewithincreasedweathering
intensity,amoderately‐weatheredsoilismorelikelytobeatamid‐developmentalstage(e.g.,
Inceptisol,Alfisol)andsupportingvegetationthatcancyclePwithoutbeingP‐depleted(e.g.,[19]).
Thismid‐CIAaccumulationcouldbereflectiveofthevegetation’simpactonsoilnutrients(e.g.,
mycorrhizalPmobilization,Precyclingthroughbiomass/organicmatterdecomposition),pointingto
akeybalancingpointinasoil’slifespanwherethebedrockisbeingweatheredenoughtomaximize
fertilitywithoutyetbeingdepleted.ForestandAltered/CultivatedsoilshavethehighestCIAvalues
(Figure7),butPconcentrationswererelativelyinvariantbetweentypesofvegetation,suggesting
thatweatheringintensityexertsagreatercontrolonsoilPthanthetypeofvegetation.
Claycontent(bothpercentofgrain‐sizefractionandabundanceofclayminerals)should
increasewiththedegreeofweatheringasoilhasundergone,andthatisreflectedintheresults(Figure
4D).Consequently,therearesecondarytrendsassociatedwithclaycontent,i.e.,claycontentpeaks
atlowerlatitudeswhereCIAvaluesarethehighest.Fetotandclaycontentshowastrongpositive
correlation,asexpected(Figure6E),butPshowsonlyaweakpotential