tree species distribution in temperate forests is …14682...tree species distribution in temperate...
TRANSCRIPT
Ecology and Evolution. 2017;7:9473–9484. | 9473www.ecolevol.org
Received:15December2016 | Revised:17August2017 | Accepted:31August2017DOI: 10.1002/ece3.3436
O R I G I N A L R E S E A R C H
Tree species distribution in temperate forests is more influenced by soil than by climate
Lorenz Walthert1 | Eliane Seraina Meier2
ThisisanopenaccessarticleunderthetermsoftheCreativeCommonsAttributionLicense,whichpermitsuse,distributionandreproductioninanymedium,providedtheoriginalworkisproperlycited.©2017TheAuthors.Ecology and EvolutionpublishedbyJohnWiley&SonsLtd.
1SwissFederalInstituteforForest,SnowandLandscapeResearchWSL,Birmensdorf,Switzerland2Agroscope,InstituteforSustainabilitySciencesISS,Zurich,Switzerland
CorrespondenceLorenzWalthert,SwissFederalInstituteforForest,SnowandLandscapeResearchWSL,ResearchUnitForestSoilsandBiogeochemistry,Birmensdorf,Switzerland.Email:[email protected]
AbstractKnowledgeof theecological requirementsdetermining tree speciesdistributions is apreconditionforsustainableforestmanagement.Atpresent,theabioticrequirementsand the relative importanceof thedifferent abiotic factors are still unclear formanytemperatetreespecies.Wetherefore investigatedtherelative importanceofclimaticandedaphicfactorsfortheabundanceof12temperatetreespeciesalongenvironmen-tal gradients. Our investigations are based on data from 1,075 forest stands acrossSwitzerlandincludingthecold-inducedtreelineofallstudiedspeciesandthedrought-inducedrangeboundariesofseveralspecies.Fourclimaticandfouredaphicpredictorsrepresentedtheimportantgrowthfactorstemperature,watersupply,nutrientavailabil-ity,andsoilaeration.Theclimaticpredictorswerederivedfromthemeteorologicalnet-workofMeteoSwiss,andtheedaphicpredictorswereavailablefromsoilprofiles.Speciescoverabundanceswererecordedinfieldsurveys.Theexplanatorypowerofthepredic-torswasassessedbyvariationpartitioninganalyseswithgeneralizedlinearmodels.Forsixofthe12species,edaphicpredictorsweremoreimportantthanclimaticpredictorsinshapingspeciesdistribution.Overallspecies,abundancesdependedmainlyonnutrientavailability, followed by temperature, water supply, and soil aeration. The often co-occurringspeciesrespondedsimilartothesegrowthfactors.Droughtturnedouttobeadeterminantofthelowerrangeboundaryforsomespecies.Weconcludethatoverall12studiedtreespecies,soilpropertiesweremoreimportantthanclimatevariablesinshap-ingtreespeciesdistribution.Theinclusionofappropriatesoilvariablesinspeciesdistri-butionmodelsallowedtobetterexplainspecies’ecologicalniches.Moreover,ourstudyrevealedthattheecologicalrequirementsoftreespeciesassessedinlocalfieldstudiesandinexperimentsarevalidatlargerscalesacrossSwitzerland.
K E Y W O R D S
drought,ecologicalniche,gradientanalysis,nutrients,soilaeration,speciesabundance,speciesdistributionmodels,variationpartitioning
1 | INTRODUCTION
Knowledgeontheecological requirementsdeterminingtreespeciesdistributions is a precondition for profitable and sustainable forest
management and for forest conservation under current and futureenvironmental conditions. In the last decades, numerous studieshavebeenconductedtodevelopandtorefinethemethodologyforassessingtreespeciesresponsestotheenvironment.Thus,areliable
9474 | WALTHERT And MEIER
conceptual framework forassessing theecological requirements fortreespeciesdistributionsshouldconsidermainly twomethodical is-sues.First,anadequatenumberofobservations(Coudun&Gegout,2006)shouldbeinvestigatedalongdiverseandstrongenvironmentalgradientsincludingspeciesrangeboundaries(Beauregard&deBlois,2014).However,speciesdonotalwaysreachtheirdistributionedgeswithintheinvestigationareaofstudies,and,thus,thispreconditionisoftennotfulfilled.Second,environmentalfactorswithadirectimpactonplantperformance(e.g.,temperatureandnutrients)shouldbepre-ferredtoindirectfactors(e.g.,elevationandgeology),becausedirectfactorsareecologicallymorecomprehensibleandhavealargerspatialapplicability (Austin & Cunningham, 1981; Guisan & Zimmermann,2000).However, thedataavailablefordirectfactorsaremostoftennot spatially explicit, for example, biotic interactions, disturbances,and soil (Mod,Scherrer, Luoto,Guisan,&Scheiner,2016).Thus, di-rect factors and, more specifically, soil characteristics are, despitetheirimportance,rarelyconsideredinpredictionsofspeciesdistribu-tions(Diekmann,Michaelis,&Pannek,2015;Thuiller,2013).Incaseswheresoil-relatedinformationhasbeenusedinstudies,ithasoftenbeenderivedfrombioindication (e.g.,Piedallu,Gegout,Lebourgeois,& Seynave, 2016) and is therefore considered to suffer from ambi-guityandthusfromlimitedcomparabilitytomeasuredsoilvariables(Szymura,Szymura,&Maciol,2014).
Only a few studies have considered measured soil variablesin species distribution models (SDM) for characterizing ecologicalniches(e.g.,Dubuisetal.,2013;Pinto&Gegout,2005;Walthert,GrafPannatier,&Meier,2013)orforpredictingspeciesdistributions(e.g.,Beauregard&deBlois,2014;Coudun,Gegout,Piedallu,&Rameau,2006; Piedallu,Gegout, Perez,& Lebourgeois, 2013).These studieshave shown that the statistical model performance was better formostofthestudiedplantspeciesifedaphicvariableswereaddedtoclimaticfactors.
FormanyEuropeantreespeciesgrowing inthetemperatezone,itisstillunclearifandhowtheyreacttosoilproperties,astheexist-ingstudiesincludedonlyafewtreespeciesordidnotfullyconsiderthemethodicalrequirementsmentionedabove;that is,observationsshouldcoverstrongenvironmentalgradientsincludingspeciesdistri-butionedges,useofdirectinsteadofindirectfactors,anduseofmea-suredsoilvariablesinsteadofecologicalindicatorvalues.
Climateandsoilareamongthemostimportantgrowthfactorsandthus drivers of tree species distributions. Important climatic factorsaretemperatureandwater,andimportantedaphicfactorsarewater,nutrients,andprobablysoilaeration.Thus,inourstudyweinvestigatethe importanceofclimatic (i.e., temperatureandwater)andedaphicfactors (i.e.,water, nutrients, and soil aeration) for the cover abun-danceof12 temperateEuropean tree species (i.e., threeconiferousandninedeciduousspecies).Morespecifically,ourstudyencompassesthethreefollowingthematicfields:First,weaimattestingforwhichtreespeciesthemodelperformancecanbe improvedbyaddingsoilvariablestoclimaticvariables,andhowmuchthespeciesdifferintheirsensitivitytoclimateandsoil.Weexpectthatincludingsoilvariablesimprovesthemodelperformanceforallspeciesandthatthecommonspecieshave the lowest sensitivity to soil properties.Moreover,we
wanttoassessthedegreeofredundancybetweentheclimateandtheedaphicvariablesinexplainingspeciesdistribution.Second,weanalyzetherelativeimportanceofthefourgrowthfactors(i.e.,temperature,water,nutrients,andsoilaeration)inshapingspeciesabundances.Weexpect similar responsepatterns for co-occurring species.Third,weevaluatespecies’sensitivitytodroughtandsoiloxygenshortage.Weexpectthatdrought isan importantfactor indeterminingthe lowerrangeboundaryofseveraltreespecies.Inaddition,wereviewwhethertheresultsofoursensitivityassessmentthatisbasedonlarge-scalein-ventorydataagreewiththeoutcomesofcorrespondingresearchfromlocalcasestudiesandfromexperiments.Asourstudyfocusesontherelative importanceof soil and climatevariables,wedonot includebioticinteractionsanddisturbances.
2 | MATERIALS AND METHODS
2.1 | Study area
ThestudyareaincludedallofSwitzerland(circa45–47°Nand6–10°E),whichislocatedinthecenterofWesternEurope(Fig.1).Duetothehighvariabilityoftopography,climate,andgeology,Switzerlandhasrelativelystrongenvironmentalgradientscomparedto itssmallsur-face.Roughly30%ofthecountry(12,000km2)iscoveredwithforest,half ofwhich is located above1,000m a.s.l. Forestmanagement ismainlypracticedatlowelevations,wherenolarge-scaleclear-cuttingis applied andnatural regeneration isoften fosteredby silviculturalmanagement(Brassel&Brändli,1999).Fertilizingorlimingtostimu-latesoilfertilityhasalwaysbeenforbiddeninSwissforests.
2.2 | Study plots, species, and environmental data
2.2.1 | Study plots
SpeciesandenvironmentaldataoriginatefromadatabaseoftheSwissFederal Institute for Forest, Snow and Landscape Research (WSL)containingdataon~1,200mainly forestedplotsacrossSwitzerland.Foreachplot,datafromafloristicinventoryconductedaccordingtoBraun-Blanquet (Braun-Blanquet,1964)andsoildatafromasoilpitwereavailable.Formostplots, thevegetation surveyand soil sam-plingwerecarriedout inthesameyearorwithamaximumtimeof5yearsbetween them.Mostplotswere selectedaccording toeco-logicalcriteriaforforestsites;thatis,speciescompositionandstandstructureshouldbeclosetothoseobservedinnaturalforests.Fromthis dataset,we excluded about 90 forest plots according to threecriteria.First,weonlyselectedmatureforeststandsbecauseweex-pectedthatthesiterequirementsoftreespeciesaremoreapparentin mature than in juvenile stands.We therefore removed all plotswith juvenile forests (n=32) from the dataset by excluding forestswherethetallesttreesweresmallerthan20m.However,maturefor-estsondrysitesorathighelevationswith limitedtreeheightwerenotexcluded.Inaddition,weexcludedallplotswherethewaterbal-ancewasnotcomputableduetounfavorablesoilproperties,mainlyboulder-richsoilswithhollowspaces.Finally,weeliminatedseveral
| 9475WALTHERT And MEIER
early successional forest plots dominated by birch trees and someheavilymanagedchestnutgroves.Theselectionprocedure resultedin1,060seminatural,predominantly latesuccessionalmature foreststands.Wefurtherincluded15treelessplotsinordertoextendthedroughtandthesoilaerationgradient.Tenoftheseplotsarelocatedindrygrassysteppesonextremelyshallowsoils,andthefiveremain-ingplotsareonmarshes.Thus,atotalof1,075plotswereinvestigatedinourstudy(Fig.1).
2.2.2 | Species data
Species data on the study plots were collected by ~35 authorsand partly originate from the Swiss Forest Vegetation Database(Wohlgemuth,2012).Theabundanceofthespeciesinthetreelayer(maturetrees)wasassessedduringthevegetationsurvey.Therefore,
alltheplantspeciesoccurringintheherb,shrub,andtreelayersinanarearangingfrom100to500m2(avg.200m2)wererecordedusingthe Braun-Blanquet cover abundance scale (Braun-Blanquet, 1964;MuellerDombois&Ellenberg,1974).Aspartofthevegetationsur-vey,theheightoftheforeststandwasestimated.Thevegetationsur-veyformostoftheplotswascarriedoutbetween1987and2014;however, 34 surveys were completed before 1987. We restrictedourassessmentto12treespeciesthatwerepresentonasufficientnumberofplotswithregardtospeciesdistributionmodels,forwhichaminimumof approximately50observations is necessary (Coudun&Gegout,2006).Threespecieswerepresent inslightly fewer than50plots (Table1).The12speciesbelong to two functionalgroups:broadleaveddeciduous (Europeanbeech,Fagus sylvatica; sycamore,Acer pseudoplatanus; European ash, Fraxinus excelsior; wych elm,Ulmus glabra; cherry tree, Prunus avium; pedunculate oak,Quercus
F IGURE 1 Locationsofthe1,075studyplotsinSwitzerland.Reddotsshowthe1,060matureforeststands,yellowsquaresthe10treelessplotsonextremelydrysoils,andbluetrianglesthefivetreelesssitesonmarshes.Forestareaisshowningreen
km
TABLE 1 Numberandelevationofforestplotswithspeciescoverabundancedataforthe12studiedtreespecies.Theminimumandmaximumelevationofallstudied1,060matureforeststandsacrossSwitzerlandwas240and2,200ma.s.l.,respectively
SpeciesNumber of plots where present
Elevation (m a.s.l.) where presentCover abundance (%) where present
Minimum Mean Maximum Mean Maximum
Picea abies 684 280 1,005 2,040 33 88
Abies alba 410 390 854 1,540 25 88
Pinus sylvestris 168 310 820 1,980 24 88
Fagus sylvatica 621 340 728 1,490 46 88
Fraxinus excelsior 206 240 613 1,360 19 63
Acer peseudoplatanus 200 290 811 1,600 17 88
Ulmus glabra 61 410 742 1,360 10 63
Prunus avium 39 240 515 820 6 38
Quercus robur 66 240 528 950 15 63
Quercus petraea 97 320 604 1,360 17 88
Quercus pubescens 41 330 727 1,360 31 88
Carpinus betulus 40 370 506 710 19 88
9476 | WALTHERT And MEIER
robur; sessile oak,Quercus petraea; downy oak,Quercus pubescens; andEuropeanhornbeam,Carpinus betulus),andneedle-leavedever-green(Norwayspruce,Picea abies;silverfir,Abies alba;andScotspine,Pinus sylvestris).Thethreeoakspecieswereidentifiedafterarevisita-tionofallplotswhereoaksoccurbyjointlyconsideringmorphologyand molecular-genetic markers according to Rellstab, Bühler, Graf,Folly,andGugerli(2016).
2.2.3 | Environmental data
Toestimatetherelativeimportanceofthesoilandclimatevariables,andtoavoidmulticollinearityproblems inouranalyses,weselectedeightpredictorsaccordingtotheirsuitabilityforexplainingtreespe-cies distributions, as described inWalthert etal. (2013). Briefly, aninitial set of 87 environmental variables was evaluated and all thevariablesthatwereselectedhadars<|0.7|.Halfoftheeightselectedpredictorswere climatic, and theotherhalfwere closely related tothesoil.
Asclimatepredictors,weselectedmeanyearlydegree-dayswitha5.56°Cthreshold(DD),meantemperatureamplitudebetweenJanuaryandJuly(T-Cont),meanamountofprecipitationfromJunetoAugust(RR),andmeanrelativeairhumidityfromJunetoAugust(RH;Table2).AirhumidityandtemperatureamplitudewerenottestedinWalthertetal.(2013),butbothhaveanimpactonplantperformanceandspe-cies distributions (for RH, see Lendzion and Leuschner (2008) andKöcher, Horna, and Leuschner (2012); forT-Cont, see Jobbagy andJackson(2000)andNishimuraandLaroque(2011)).
Climate data for the period 1981–2010 were derived fromweather data from themeteorological network ofMeteoSwiss.DDvalues were produced with Daymet (Thornton, Running, & White,1997).T-Cont,RR,andRHwerecalculatedusingtherespectivelong-termmeanmonthlyvalues(1981–2010)providedbyRemund,Rihm,andHuguenin-Landl(2014).
Assoilpredictors,weselectedthedroughtindex(AT/PT)forsoilwater availability, theC/N ratio (C/N), and base saturation (BS) forsoilnutrientavailability,and thedepthof thewater level in thesoil(W-level)forsoilaeration(Table2).Themethodsforassessingthesesoilpredictorsarebrieflydescribedinthefollowingtext.Foradetaileddescriptionofthesepredictorsanda justificationfortheirselection,seeAppendixS1.
Soilsamplesweretakenfromsoilprofiles,onaverage1.2mdeep,accordingtopedogenetichorizons.
BS in0–50cmsoildepth:Exchangeablecations (Na,K,Mg,Ca,Mn,Al,Fe)wereextractedina1mol/LNH4Clsolutionanddeterminedbyatomicemissionspectroscopy(ICP-AES).Contentsofexchangeableprotonswere determined after extracting the soil in a 1mol/LKCLsolution.Theeffectivecation-exchangecapacity(CEC)wascalculatedbysumming thechargeequivalentsofexchangeableNa,K,Mg,Ca,Mn,Al,Fe,andH.ThebasesaturationisthepercentofexchangeableNa,K,Mg,andCaoftheCEC.
C/Nin0–10cmsoildepth:C/Nistheratiobetweenorganiccar-bonandtotalnitrogen.Theircontentsweredeterminedbydrycom-bustion.PossiblecarbonateswereremovedbyHClvaporpriortodrycombustion.
AT/PT: This index represents the ratio of actual to potentialtranspiration and corresponds to the average reduction in tran-spiration due to soilwater shortage from June toAugust in theperiod1981–2010.Thewaterbalancewasmodeledonall1,075studyplotsusingacoupledmassandheattransfermodelforsoil–plant–atmosphere systems (Coupmodel). The model was drivenby daily weather data and used soil hydraulic parameters thatwe derived from measured soil properties like texture, density,andstonecontent.Vegetationwasimplementedasforadynamicmodel forest,andmaximumrootingdepthwasset to1.5m.Dueto relatively high amounts of precipitation (>900mm/a) in mostpartsofSwitzerland,droughtpredominantlyoccursonsoilswitha
TABLE 2 Environmentalpredictorsatthe1,075studyplotsacrossSwitzerland
Category Name Class Description Min Q0.25 Med Q0.75 Max Unit
Climate DD Temperature Meanyearlydegree-days>5.56°C(1981–2010)
575 1,318 1,743 2,025 2,870 °C
T-Cont Temperature MeantemperatureamplitudeJanuary/July(1981–2010)
14.6 17.1 18.0 18.4 20.5 °C
RR Water MeanprecipitationJunetoAugust(1981–2010)
151 342 414 510 792 mm
RH Water MeanrelativeairhumidityJunetoAugust(1981–2010)
63.7 71.8 73.0 74.3 80.9 %
Soil AT/PT Water Droughtindex;meanratiobetweenactualandpotentialtranspirationJunetoAugust(1981–2010)
0.23 0.94 0.99 1.00 1.00 —
C/N Nutrients MeanC/Nratio(Corg/Ntot)in0–10cmsoildepth
7.5 14.5 16.9 20.1 42.5 —
BS Nutrients Meanbasesaturationin0–50cmsoildepth
2.7 18.5 85.9 99.6 100.0 %
W-level Soilaeration Meandepthofthesoilwaterlevelinthevegetationperiod
20 200 200 200 200 cm
| 9477WALTHERT And MEIER
lowstoragecapacityofplantavailablewater.Inthehumidsuboce-anicclimateofcentralEurope,edaphicfactors,suchaslowwaterholdingcapacity,areexpectedtoimposemoreseriousconstraintson treewater relations thanclimate (Backes&Leuschner,2000).Therefore,we assigned the drought indexAT/PT to the edaphicpredictors.Thus, inourstudy,wateravailabilityisrepresentedbytwo climatic predictors (RR and RH) and one edaphic predictor (AT/PT).
W-Level:Thisvariablewasusedasaproxyforsoiloxygenshort-age.Itisderivedfromredoximorphicpropertiesandtimeseriesofthemeasuredsoilwaterlevelinthesoilpits.
2.3 | Data analysis
Toestimatetherelativeimportanceofclimate(temperature[DD,T-Cont] and water [RR, RH]) and soil (water [AT/PT], nutrients[C/N,BS],andaeration[W-level]),weconductedavariationpar-titioning analysis (Borcard, Legendre, & Drapeau, 1992; Mood,1971).Weestimated thepurecontributionofeachpredictor setby subtracting the model fit of the opposite set of predictorsfrom the full setofpredictors, so thatVPredictor subset i = VFullModel–VFullModelwithoutPredictorsubseti.Asstatisticalmodels,weusedgener-alized linearmodels (GLM,McCullagh&Nelder,1989)with logitlinks,assumingabinomialdistributionbecausetheresponsevari-ableswere proportions (i.e., species abundance data).Model fitswereevaluatedusingtheadjustedD2(adj.D2),followingWeisberg(1980).ResponsecurveswerederivedforeachtreespeciesfrompredictionsofGLMswithlogitlinksassumingabinomialdistribu-tion.Asresponsevariable,weusedthecoverabundanceofmaturetreesandaspredictorsthesinglevariablesinthelinearandquad-raticforms.ToestimatetheGLMperformanceforeachcombina-tionofsinglepredictorandspecies,weestimatedwiththehelpofanANOVAthesignificanceofthedifferencebetweentheperfor-mancesofthefullmodelandthemodelwithoutthespecificsinglepredictor.AlldatawerepreparedandanalyzedusingR (R,2016)andArcGIS9.2(ESRI,2006).
3 | RESULTS
3.1 | Relative importance of climate and soil
Thestatisticalperformance (model fits)of the fullmodels includ-ing all climatic and edaphic variables varied considerably amongtreespecies.Theexplaineddeviance(adj.D2) rangedfrom0.28to0.59(Table3).Thebestmodelfits (i.e.,adj.D2>0.45)werefoundfor P. sylvestris, F. excelsior, and Q. pubescens. Poor model fits(i.e.,adj.D2<0.30)were foundforP. abies,A. pseudoplatanus, andA. alba.
Across all 12 tree species, the eight single predictors (in total96 combinations) had a significant amount of information for thedistribution of the tree species in 15 cases (significance level ofp<.05;Table3).Formostspecies,p-valuesandadj.D2werehighlycorrelated.
Foralltreespecies,modelsincludingedaphicandclimaticpredic-torshadabettermodel fit thanmodels includingonlyclimaticpre-dictors (Table4).Further,forallspecies,themeanpurecontributionof the edaphic predictors (adj.D2 0.13±0.09) was larger than thecorresponding pure contribution of the climatic predictors (adj.D2 0.11±0.06).However,thepurecontributionsoftheedaphicpredic-torsvaried strongly among the tree species: Forhalf of the species(F. excelsior,A. pseudoplatanus,U. glabra,Q. pubescens,P. sylvestris,andQ. petraea),thepurecontributionoftheedaphicpredictorswaslarger,while for the other half of the species (A. alba,F. sylvatica,Q. robur,C. betulus, P. avium, and P. abies), the pure contribution of the cli-maticpredictorswaslarger(Fig.2).ThemostcommontreespeciesinSwitzerland,P. abies,A. alba,andF. sylvaticarespondedratherweaklytosoilcharacteristics.
Acrossall species, the jointcontribution (redundancy)ofclimateandsoil (adj.D20.15±0.10;Table4)wassimilar to thepureclimate(adj.D20.11±0.06)andthepuresoilcontribution(adj.D2 0.13 ± 0.09). Redundancyofsoilandclimatevariableswasobservedforallspecies,andthedegreeofthisredundancy,however,variedbetweenspecies.
3.2 | Relative importance of temperature, water, nutrients, and soil aeration
Themeanvalueof thepurecontributionsofall12speciesshowedthat the abundance depended mainly on nutrient status (adj.D2 0.09±0.09), followed by temperature (adj.D2 0.07±0.05), wateravailability(adj.D20.04±0.04),andsoilaeration(adj.D2 0.01 ± 0.01; Table4).Basedon the individual responsepatterns, the12 speciesweregroupedasbeingmainlysensitivetonutrientstatus,totempera-tureortowateravailability(Fig.3).
Species with a predominant response to nutrient status wereA. pseudoplatanus, U. glabra, F. excelsior, P. avium, and P. sylvestris (Fig.3).AbundanceofA. pseudoplatanus, U. glabra, F. excelsior,and,toalesserextent,P. aviumwaspositivelycorrelatedwithnutrientavail-ability.These speciespreferred soilswithhighbase saturations andlowC/Nratios(Table3).WhileP. sylvestrisalsorespondedpositivelytohighbasesaturations,thisspeciesreachedthehighestabundanceon soilswith low nitrogen availabilities (high C/N ratios).After nu-trient status, temperature, andwater availabilitywere of secondaryimportanceforthesefivespecies.WhileA. pseudoplatanusrespondedonlymarginallytotemperature,thefourotherspeciesrequiredhighertemperatures.Highairhumidityandagoodwatersupplywereben-eficial forA. pseudoplatanusandforU. glabra. Incontrast,F. excelsior, P. avium,and P. sylvestriswerelesswaterdemanding(Table3;seealsoSection3.3.1).
Species that predominantly responded to temperature wereQ. robur, C. betulus, F. sylvatica, A. alba,andP. abies (Fig.3).Whiletheabundance of Q. robur and C. betulus increased with temperature,F. sylvatica, A. alba, and P. abies reached the highest abundance atmoderatetemperatures; the lowestheatrequirementwasfoundforP. abieswithamaximalabundanceat1,100°Cdegree-days,followedby A. alba and then F. sylvatica with corresponding degree-days of1,500°Cand1,900°C,respectively(Table3).Aftertemperature,water
9478 | WALTHERT And MEIER
andnutrientstatuswereofsecondaryimportanceforthesefivespe-cies.P. abies, A. alba, and F. sylvatica had higherwater requirementsthan Q. robur and C. betulus (Table3; see also Section3.3.1), withF. sylvaticaandA. albabenefitingmostfromhighairhumidity.Allfivespecieswere rather indifferent to soil nutrient status,most notablywithrespecttobasesaturation(Table3).
ApredominantresponsetowateravailabilitywasfoundforQ. pu-bescens andQ. petraea (Fig.3).Bothspeciesweremostabundantatsiteswherewatersupplyandairhumiditywereratherlow(Table3;seealsoSection3.3.1).Moreover,bothspeciespreferredrelativelywarmsites.Thenutrientstatusofthesoilhadonlyasmallinfluenceontheseoakspecies.
3.3 | Responses of species abundance along drought and soil aeration gradients
3.3.1 | Drought sensitivity
Amongthethreewater-relatedpredictorsofourstudy(RR,RH,andAT/PT),thedroughtindexAT/PTwasmostabletoexplainspeciescoverabundances;thatis,thepurecontributionofthisvariablewasrelativelyhigh for themajorityof thespecieswithsignificantvalues,however,onlyforP. abies,F. sylvatica,andQ. pubescens(p<.05;Table3).
The12specieswerearrangedinfivegroupsaccordingtotheirre-sponsetoAT/PT.Thefirstgroup,withP. abies, A. alba,andF. sylvatica,
hadamaximalabundanceundergoodwatersupply, that is,athigh AT/PT levels, and faded away at AT/PT values of approximately 0.6–0.7(Fig.4).Thesecondgroup,withA. pseudoplatanusandU. glabra,behaved similar to the first group, but hadmuch lower abundancelevels, as is typical for these tree species that rarely dominate inSwiss forests (Table1). Compared to the speciesmentioned above,the abundance of C. betulus and Q. robur peaked on drier sites (AT/PT 0.8). The species of group 4,Q. pubescens, P. sylvestris, andQ. petraeapeakedinabundanceinwhatwasclearlythedriestrange(AT/PT 0.6–0.7) and all extendedmost toward the dry end of thedroughtaxis.Finally,auniqueresponsewasfoundforF. excelsior.Thisspecies,thoughoftenpresentonhydromorphicsoils,extendedintotheverydryrangeofthedroughtaxis.
BasedontheirresponsetothedroughtindexAT/PT,that is,thepositionoftheabundancemaximumandtheshapeofthecurveto-ward thedryendof this gradient (Fig.4), thedrought sensitivityofthestudiedspecieswasrankedasfollows:U. glabra, A. pseudoplatanus, F. sylvatica, P. abies, A. alba > C. betulus, Q. robur > P. avium, Q. petraea, P. sylvestris > F. excelsior, Q. pubescens.
3.3.2 | Sensitivity to limited soil aeration
Asmeasuredbythepurecontributions,soilaerationhadlittleinfluenceonthecoverabundanceofmosttreespecies(Table3).ForF. sylvatica,however,thepredictiveabilityofthisvariablewasquitehigh(p < .001).
TABLE 3 Purecontributionsofindividualpredictorsandofthefullsetofpredictors(fullmodel)forthe12studiedtreespecies,derived fromvariableGLMswithcoverabundanceofmaturetreesastheresponsevariableandenvironmentalpredictorsbasedondatafrom1,075 studyplotsacrossSwitzerland.Numbersindicatethedevianceexplainedbytheindividualpredictors(adj.D2multipliedby100)andbythefull model(adj.D2).Thedirectionofthetrend(T)betweenthepredictorandtheresponsevariableisindicatedaspositivelinear“+,” negativelinear“−,”positiveunimodal“+/−,”orwithnocleartrend“N.”“Optimum”(Opt.)forDD,RH,AT/PT,andC/Nspecifiesthe environmentalconditionsforwhichmaximalabundancewaspredictedifthetrendwaspositiveunimodal.Boldfaceindicatesadj.D2≥1.5%for individualpredictors.Forp-values,see(p).AbbreviationsandunitsofpredictorsareexplainedinTable2.Notethatthesumofthepure contributionsoftheindividualpredictorsdiffersfromthedevianceexplainedbythefullmodelduetojointcontributions
Species
Full model
Climate Soil
Temperature Water Nutrients Soil aeration
Full modelDD TOpt. °C p T- Cont T p RR T p RH T
Opt. % p AT/PT T Opt. p C/N T Opt. p BS T p W- Level T p
P. abies 5.70 +/− 1100 .00 0.10 − .64 −0.10 + .95 0.00 + .76 1.60 + .02 1.70 +/− 27 .02 0.50 − .26 0.40 − .32 0.28
A. alba 5.50 +/− 1500 .00 0.00 +/− .86 0.40 + .40 3.30 +/− 76 .00 0.70 + .28 0.60 +/− 18 .34 0.60 − .31 0.60 +/− .29 0.30
P. sylvestris 2.60 +/− 1900 .06 0.60 + .47 −0.10 N .95 1.20 − .28 0.50 +/− 0.65 .53 9.50 + .00 2.60 + .07 0.50 + .54 0.45
F. sylvatica 10.10 +/− 1900 .00 1.20 +/− .01 0.30 + .29 0.30 +/− 73 .27 1.90 +/− 0.90 .00 1.40 +/− 15 .01 0.50 N .12 3.30 + .00 0.37
F. excelsior 2.60 + .09 0.20 +/− .77 0.00 − .90 0.40 − .62 0.70 − .48 8.00 − .00 2.80 + .07 1.10 − .33 0.55
A. ps.platanus −0.10 N .97 −0.10 − .98 0.20 + .73 1.60 +/− 75 .24 0.20 + .74 7.30 − .00 2.50 + .11 1.10 N .36 0.29
U. glabra 4.80 + .35 3.20 N .49 0.20 + .94 2.10 +/− 74 .63 4.30 + .39 1.50 − .71 7.20 + .21 3.50 N .46 0.39
P. avium 4.00 + .66 1.40 + .86 0.70 N .92 1.60 − .84 0.80 − .92 2.90 − .74 0.70 N .92 1.40 N .86 0.39
Q. robur 8.50 + .04 0.80 +/− .71 1.00 N .66 0.60 − .78 1.90 +/− 0.80 .47 0.60 − .78 0.70 N .73 0.20 N .90 0.40
Q. petraea 1.90 + .35 0.10 + .88 1.00 N .57 1.90 +/− 68 .34 4.00 +/− 0.70 .12 1.50 +/− 19 .42 0.40 N .75 0.50 + .72 0.30
Q. pubescens 0.60 + .72 1.20 + .52 1.40 N .49 2.50 − .27 9.70 +/− 0.60 .01 2.50 +/− 18 .28 0.20 + .87 0.20 + .89 0.59
C. betulus 3.20 + .36 0.20 +/− .92 4.50 N .24 0.90 − .72 2.20 +/− 0.80 .49 2.20 − .48 0.10 N .92 0.40 N .86 0.40
| 9479WALTHERT And MEIER
Ourdata enabledus to assess the sensitivity to limited soil aerationforfivespecies.Basedontheshapeoftheresponsecurves,werankedF. sylvaticaassensitiveandP. abies,A. alba,F. excelsior,andA. pseudopla-tanusasnotsensitivetolimitedsoilaeration(Fig.5).AswedidnothaveenoughobservationsonstronglyhydromorphicsoilsforP. sylvestrisandforQ. petraea,wedidnotassessthesensitivitytosoilaerationforthesespecies.ThesamewasthecaseforC. betulus, U. glabra, P. avium, Q. pu-bescens,andQ. robur,whichhadnearlylinearandhorizontalresponsecurvesatlowabundancelevelsandwerethusnotevaluable.
4 | DISCUSSION
4.1 | Relative importance of climate and soil
Variationpartitioningrevealedthattheadditionofedaphicvariablestotheclimaticvariablesimprovedthemodelperformanceforallspe-cies.Similar resultswerealready found inearlier research,both fortreespecies(e.g.,Beauregard&deBlois,2014;Coudunetal.,2006;Piedalluetal.,2013)and forherbaceousspecies (e.g.,Dubuisetal.,2013).Overallspecies,soilpropertieswereevenmoreimportantthanclimate variables in explaining species distribution. Diekmann etal.(2015) found that thedistributionofmanyplant species is stronglydrivenbysoilconditions inregionswith lowclimaticheterogeneity.Ourstudyshowsthatthisseemstobethecasenotonlyinclimaticallyhomogeneous landscapesbut alsoeven inheterogeneousones like
thoseinSwitzerland.Therefore,appropriatesoilpropertiesshouldbeincludedinspeciesdistributionmodels,especiallyforthespecieswithastrongresponsetoedaphicvariables.Forthemostcommonspecies,P. abies,A. alba,andF. sylvatica,however,theinclusionofsoilvariablesislesseffective,astheirwidespreaddistributioncanbeinterpretedasaresultoftheirrelativelylowsensitivitytosoilcharacteristics.
Climateisasoil-formingfactoraffectingnumeroussoilprocessesandsoilproperties,whichcouldresult inacertaindegreeof redun-dancybetweensoilandclimatevariablesinexplainingspeciesdistri-bution,asrecentlyhypothesizedbyThuiller(2013).Afirstindicationof such a redundancy in our dataset is given by the relatively highcorrelationbetween the four climate and the four soil predictors insomecases(Pearsonrrangingfrom−0.40to0.55forthe16combi-nations).Duetothesepartlysubstantialcorrelations,theredundancyinexplainingspeciesdistribution,measuredasjointcontributionofallclimateandallsoilvariablesoverallspecies,wasofsimilarmagnitudethanthepureclimateandthepuresoilcontributionsandthus,con-firmingthehypothesisofThuiller(2013).
Themodelperformancewiththefullsetofpredictorswasratherweak for some species. Part of the unexplained deviance could beattributed tobiotic factorsnot included inourmodels (Meieretal.,2010).Finally, it isnoteworthythattheresponseswereall inaccor-dancewith theniche theory (Austin&Smith,1990),withnomulti-modalorU-shapedresponsecurvesbutonlylineartrendsorunimodalorskewedresponses(Table3andFig.4).
TABLE 3 Purecontributionsofindividualpredictorsandofthefullsetofpredictors(fullmodel)forthe12studiedtreespecies,derived fromvariableGLMswithcoverabundanceofmaturetreesastheresponsevariableandenvironmentalpredictorsbasedondatafrom1,075 studyplotsacrossSwitzerland.Numbersindicatethedevianceexplainedbytheindividualpredictors(adj.D2multipliedby100)andbythefull model(adj.D2).Thedirectionofthetrend(T)betweenthepredictorandtheresponsevariableisindicatedaspositivelinear“+,” negativelinear“−,”positiveunimodal“+/−,”orwithnocleartrend“N.”“Optimum”(Opt.)forDD,RH,AT/PT,andC/Nspecifiesthe environmentalconditionsforwhichmaximalabundancewaspredictedifthetrendwaspositiveunimodal.Boldfaceindicatesadj.D2≥1.5%for individualpredictors.Forp-values,see(p).AbbreviationsandunitsofpredictorsareexplainedinTable2.Notethatthesumofthepure contributionsoftheindividualpredictorsdiffersfromthedevianceexplainedbythefullmodelduetojointcontributions
Species
Full model
Climate Soil
Temperature Water Nutrients Soil aeration
Full modelDD TOpt. °C p T- Cont T p RR T p RH T
Opt. % p AT/PT T Opt. p C/N T Opt. p BS T p W- Level T p
P. abies 5.70 +/− 1100 .00 0.10 − .64 −0.10 + .95 0.00 + .76 1.60 + .02 1.70 +/− 27 .02 0.50 − .26 0.40 − .32 0.28
A. alba 5.50 +/− 1500 .00 0.00 +/− .86 0.40 + .40 3.30 +/− 76 .00 0.70 + .28 0.60 +/− 18 .34 0.60 − .31 0.60 +/− .29 0.30
P. sylvestris 2.60 +/− 1900 .06 0.60 + .47 −0.10 N .95 1.20 − .28 0.50 +/− 0.65 .53 9.50 + .00 2.60 + .07 0.50 + .54 0.45
F. sylvatica 10.10 +/− 1900 .00 1.20 +/− .01 0.30 + .29 0.30 +/− 73 .27 1.90 +/− 0.90 .00 1.40 +/− 15 .01 0.50 N .12 3.30 + .00 0.37
F. excelsior 2.60 + .09 0.20 +/− .77 0.00 − .90 0.40 − .62 0.70 − .48 8.00 − .00 2.80 + .07 1.10 − .33 0.55
A. ps.platanus −0.10 N .97 −0.10 − .98 0.20 + .73 1.60 +/− 75 .24 0.20 + .74 7.30 − .00 2.50 + .11 1.10 N .36 0.29
U. glabra 4.80 + .35 3.20 N .49 0.20 + .94 2.10 +/− 74 .63 4.30 + .39 1.50 − .71 7.20 + .21 3.50 N .46 0.39
P. avium 4.00 + .66 1.40 + .86 0.70 N .92 1.60 − .84 0.80 − .92 2.90 − .74 0.70 N .92 1.40 N .86 0.39
Q. robur 8.50 + .04 0.80 +/− .71 1.00 N .66 0.60 − .78 1.90 +/− 0.80 .47 0.60 − .78 0.70 N .73 0.20 N .90 0.40
Q. petraea 1.90 + .35 0.10 + .88 1.00 N .57 1.90 +/− 68 .34 4.00 +/− 0.70 .12 1.50 +/− 19 .42 0.40 N .75 0.50 + .72 0.30
Q. pubescens 0.60 + .72 1.20 + .52 1.40 N .49 2.50 − .27 9.70 +/− 0.60 .01 2.50 +/− 18 .28 0.20 + .87 0.20 + .89 0.59
C. betulus 3.20 + .36 0.20 +/− .92 4.50 N .24 0.90 − .72 2.20 +/− 0.80 .49 2.20 − .48 0.10 N .92 0.40 N .86 0.40
9480 | WALTHERT And MEIER
4.2 | Relative importance of temperature, water, nutrients, and soil aeration
Asweexpected, theoftenco-occurring treespeciesshowedsimi-lar response patterns to the four growth factors. Accordingly, thefour species F. excelsior, A. pseudoplatanus, U. glabra, and P. avium reached highest abundances on nutrient-rich soils,Q. petraea andQ. pubescensweremostabundantonsiteswitha lowwateravail-ability,whereasQ. roburandC. betuluswererestrictedtowarmsiteswith frequently changing soil water and soil oxygen availabilities.Thethreemostwidespreadspecies,P. abies,A. alba,andF. sylvatica,showed a similar response pattern as well. The characteristic forthese specieswasa lownutrientdemandanda relativelyhigh re-sponsetotemperature.
Acrossall12species, temperatureandnutrientavailabilityweremost important in shaping species ecological niches. Consistentlywiththealtitudinaldistributionofthestudiedtreespecies,variationpartitioningrevealedthedifferentheatdemand(degree-days)ofthespecies (e.g., F. sylvatica>A. alba>P. abies; Q. robur>Q. petraea). Temperaturewasa relevant factornotonly inour studybutalso inmostotherSDMstudies(Austin&VanNiel,2011).Contrarytodegree-days,thermalcontinentalitywasaweakpredictorformostspeciesinour study.The impact of soil nutrient availability on tree growth isspeciesspecific(Lévesque,Walthert,&Weber,2016).Therefore,itis
logicalthatspeciesdifferedintheirresponsetonutrientsinourstudy.However, there isanongoingdebateabout thedegreeof feedbackbetween plants and soils. The soil does not only influence speciescomposition,but speciesalsoaffect soilproperties suchasnutrientstatus, mainly by species specific litter input. This bidirectional linkbetween plants and soils is discussed inmore detail in Section4.4.Afternutrientstatusandtemperature,wateravailabilitywasalsoanimportantfactorforspeciesnichedifferentiation,muchmoresothansoil aeration. Among the three predictors that represented wateravailability,thedroughtindexAT/PTwasbyfarmostimportant(seeSection4.3.1),followedbyairhumidityandprecipitation,whichwasaweakpredictorformostspecies.Speciesthatpreferredoceaniccli-matewithhighmeanairhumiditywereA. alba,A. pseudoplatanusandU. glabra,whereasQ. pubescens,P. sylvestris, andP. aviumweremostabundantincontinentalregionswithoftenrelativelylowairhumidity.Theimportanceofsoilaerationcouldbeevaluatedonlyforaminorityofthestudiedspecies(seeSection4.3.2).
4.3 | Responses of species abundance along drought and soil aeration gradients
Futuresummerswillprobablybewarmeranddrier,whileprecipitationinwinterandspringmaybeenhanced(Lindneretal.,2014).Increasing
F IGURE 3 Relativeimportanceoftemperature,water,nutrients,andsoilaerationinexplainingthedistributionofthe12studiedtreespecies.RelativeimportanceisbasedonpurecontributionsderivedfromgroupedvariableGLMswithcoverabundanceofmaturetreesastheresponsevariableandfourgroupedenvironmentalpredictors—temperature,water,nutrients,andsoilaeration—basedondatafrom1,075studyplotsacrossSwitzerland.Togetrelativeimportance,thepurecontributionsofthefourgroupedpredictorswereconvertedtorelativeonessothatthesumofthepurecontributionsofthefourgroupedpredictorswassetto100%foreachspecies.Forpurecontributions(adj.D2)ofthefourgroupedpredictorsandofthefullmodels,seeTable4
0
20
40
60
80
100
Ace.p
se
Fra.e
xc
Pin.s
yl Ul
m.gla
Pr
u.avi
Que.r
ob
Fag.s
yl Ab
i.alb
Pic.a
bi Ca
r.bet
Que.p
ub
Que.p
et
Relat
ive im
porta
nce (
%)
Nutrientdominated
Temperaturedominated
Waterdominated
Nutrients Soil aerationWaterTemperature
F IGURE 2 Climateversussoilsensitivityforthe12studiedtreespecies.AbsoluteandrelativepurecontributionsofclimateandsoilvariableswerederivedfromgroupedvariableGLMswithcoverabundanceofmaturetreesastheresponsevariableandgroupedenvironmentalpredictors(fourclimateandfoursoilpredictors)basedondatafrom1,075studyplotsacrossSwitzerland.Speciesbelowthe1:1lineareconsideredrathersensitivetoclimate,whereasspeciesabovethislineareconsideredrathersensitivetosoilvariables.Therelativepurecontributionsrefertothefullmodel
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Soil v
ariab
les
Climate variables
Fag.syl
Absolute pure contributionRelative pure contribution
Abi.albQue.rob
Car.bet
Pru.aviPic.abi
Que.petPin.syl
Fra.excUlm.gla
Ace.pse
Que.pub
| 9481WALTHERT And MEIER
summerdroughtsanddecreasingoxygenavailabilityduetowaterlog-gingduringwinter and springmayadversely affect thegrowthandcompetitive ability of sensitive tree species on many forest sites
(Kreuzwieser&Gessler,2010).Withthesefutureprospects,itisim-portanttoknowthesensitivityofspeciestodroughtandtosoiloxy-genshortage.
4.3.1 | Drought
Almost every physiological process in plants is affected directly orindirectlybywater supply (Kramer&Boyer,1995). In thepast, thedroughttoleranceofmanytreespecieshasbeenthoroughlyexplored,mainlyonthebasisofcasestudies.However, it isunclearhowwelltheselocalassessmentsreflectthelarge-scalesituationinforesteco-systems.Moreover,previousstudieshaveoftenfocusedonjuvenileplants,althoughjuvenileandmatureplantsmaydiffer inphysiology(Kolb&Matyssek,2001;Ryan&Yoder,1997).Furthermore,growthconditions simulated in experiments diverge from themultifactorialfieldconditions(Kolb&Matyssek,2001).
Basedonlarge-scaleinventorydatafromabout1,000maturefor-eststands,weshowedthatthe12studiedtreespeciesvariedgreatlyin theirdrought sensitivity, that is, in their response to thesummerdrought indexAT/PT.Our results are ingoodqualitativeagreementwithcomparableresearchexploringspeciesdroughtsensitivity,suchas (1) case studies investigating physiological characteristics ofma-turetrees inCentralEuropeanforeststands (e.g.,Zweifel,Rigling,&Dobbertin, 2009) or juvenile trees in greenhouse experiments (e.g.,Arend,Brem,Kuster,&Gunthardt-Goerg,2013);(2)aSwisscasestudymeasuringthecanopyfoliagetemperatureofmaturetreeswithinfra-redthermography(Scherrer,Bader,&Körner,2011);(3)casestudiesongrowthdynamicsofmaturetreesinold-growthEuropeanforestsusingtreerings(e.g.,Cavin,Mountford,Peterken,&Jump,2013);and(4)large-scaletreeregenerationandmortalityassessmentsinmature
F IGURE 4 Speciessensitivitytodrought.ResponsecurveswerederivedfromsinglevariableGLMswithcoverabundanceofmaturetreesastheresponsevariableandadroughtindex(meanAT/PTbetweenJuneandAugust,1981–2010)aspredictorbasedondatafrom1,075studyplotsacrossSwitzerland.Thedifferentspeciesabundancemaximareflectthedifferentmeanabundancethatspeciesreachedonthestudyplots(Table1).Prunus aviumisnotillustratedbecauseitsresponsecurverunshorizontallyatalevelofabout0%abundance.IfmodeledAT/PTissmallerthanabout0.3–0.4,forestsarenotabletopersistduetoexcessivewatershortage
0.4
0.3
0.2
0.1
0.0
0.4 0.5 0.6 0.7 0.8 0.9 1.0
Pred
icted
abun
danc
e
Drought index AT/PT during summerDry Moist
Fag.syl
Abi.al
b
Fra.exc
Pic.ab
iPin.syl
Que.pet
Que.pub
Que.rob
Car.bet
Ulm.glaAce.pse
TABLE 4 Pureandjointcontributionsofgroupedpredictorsforthe12studiedspecies,derivedfromgroupedvariableGLMswithcoverabundanceofmaturetreesastheresponsevariableandenvironmentalpredictorsbasedondatafrom1,075studyplotsacrossSwitzerland.Numbersindicatethedevianceexplainedbythegroupedpredictors(adj.D2).Notethatthesumofthepurecontributionsofthegroupedpredictorsdiffersfromthedevianceexplainedbythefullmodelduetojointcontributions
Species
Pure contributions adj.D2
Joint contribu-tions adj.D2
Full model Climate Soil Temperature Water Nutrients S. aeration Climate and soil
P. abies 0.28 0.08 0.07 0.06 0.02 0.04 0.00 0.13
A. alba 0.30 0.15 0.04 0.09 0.05 0.02 0.01 0.10
P. sylvestris 0.45 0.07 0.13 0.04 0.02 0.10 0.01 0.24
F. sylvatica 0.37 0.17 0.08 0.14 0.03 0.02 0.03 0.11
F. excelsior 0.55 0.12 0.34 0.07 0.01 0.29 0.01 0.09
A. ps.platanus 0.29 0.02 0.21 0.00 0.03 0.21 0.01 0.06
U. glabra 0.39 0.08 0.24 0.05 0.08 0.21 0.03 0.07
P. avium 0.39 0.14 0.11 0.07 0.03 0.08 0.01 0.14
Q. robur 0.40 0.19 0.05 0.17 0.02 0.03 0.00 0.15
Q. petraea 0.30 0.06 0.09 0.02 0.08 0.04 0.00 0.16
Q. pubescens 0.59 0.05 0.13 0.02 0.14 0.03 0.00 0.41
C. betulus 0.40 0.16 0.07 0.05 0.03 0.05 0.00 0.16
9482 | WALTHERT And MEIER
Swissforests(Riglingetal.,2013).However,thecomparisonbetweenourstudyandtheabove-mentioneddrought-orientedresearchisnotexhaustive.First,comparabledatawerehardlyavailableforsomeofthe 12 studied species, especially forU. glabra, P. avium, andC. bet-ulus. Second,withonly three speciesonaverage, thenumberof si-multaneously investigated specieswas rather low inmost of thesepreviousstudiesfocusedondrought.Nevertheless,ourstudybasedonlarge-scaleinventorydataindicatesthatspeciesdroughtsensitivi-ties,asassessedincasestudiesandexperiments,arewellreflectedinspeciescoverabundanceinmatureforeststandsacrossSwitzerland.Moreover,species’responsepatternstothedroughtindexAT/PTin-dicatedthatdroughtwasanimportantdeterminantofthelowerrangeboundaryforsomespecies,ashypothesizedbyNormandetal.(2009)andAndereggandHilleRisLambers(2016).
4.3.2 | Soil aeration
Treesareaerobicorganisms thatdependonasteadysupplyofox-ygen toall livingcells.A shortageofoxygen thereforedisturbs theplant metabolism. During waterlogging, oxygen diffusion into thesoil andoxygen supply to the roots are reduced (Kozlowski, 1986).Manyenergy-consumingprocesses,includingshootandrootgrowth,aresloweddowntoovercometheenergycrisiscausedbywaterlog-ging (Kreuzwieser&Rennenberg,2014).Currentknowledgeon theresponses of European tree species to waterlogging and oxygen
shortagehasbeenacquiredmainlyincasestudiesbasedon(1)physi-ological, morphological, and growth reactions of seedlings undercontrolledconditions (e.g.,Dreyer,1994;Schmull&Thomas,2000),(2)abovegrounddamagetomaturetrees inopen land (reviewfromKreuzwieser&Rennenberg, 2014), and (3) time until plant diebackoccurred(reviewfromNiinemets&Valladares,2006).Ingeneral,spe-ciescoverabundanceontheapproximately1,000studyplotsacrossSwitzerlandreflectedspeciessensitivityratingsfromthesecasestud-ies.However, as alreadyobserved for drought sensitivity, compari-sonsbetweenpublishedresearchandourstudysufferedfromscarcedataavailability.Moreover,wecouldcompareonlyfivespecies(F. syl-vatica, P. abies, A. alba, F. excelsior,andA. pseudoplatanus),asoursen-sitivityassessmentwasfeasibleonlyforthesespecies.Discrepanciesbetween our and published resultswere only found in the case ofNiinemetsandValladares(2006),whostatedthatA. alba, P. abies,andA. pseudoplatanusareveryintoleranttowaterlogging.Inourstudy,incontrast,thesespeciesreachedarelativelyhighabundanceonmanystronglyhydromorphicsoils.TheabilityofP. abiestoadapttowater-loggingisremarkable.Instronglyhydromorphicsoils,thisspeciesde-velopsashallowrootingsystemtoavoiddeeperanaerobichorizons.
4.4 | Feedback between plants and soils
Thepostulatedrelevanceofsoilvariablesfortreespeciesdistribu-tionmayarisefromcircularreasoning.Vegetationisasoil-formingfactorandthusmodifiesdifferentsoilcharacteristicsmainlyinthetopsoil (Augusto, Dupouey, & Ranger, 2003). Even though suchtopsoilmodificationsmaybeconsiderableandmayoccurrelativelyfastwithinthelifetimeofatreespecies(Cools,Vesterdal,DeVos,Vanguelova,&Hansen,2014;Vesterdal,Schmidt,Callesen,Nilsson,&Gundersen, 2008), the influence of tree species on soil shouldnotbeoverratedasgeologicalcharacteristicsandothersitefactorsseemto influence soilpropertiesat leastasmuchas tree speciesdo (Augusto etal., 2003; Prescott, 2002). Therefore, the appar-entcircularreasoninginourstudymaybeonlyweakmeaningthatsoil datamaywell beusedas variables for analyzing tree speciesdistribution.
4.5 | Importance of forest management
Silvicultural interventionshavethepotentialtosmeartheecologicalniches of tree species. Past forest management probably changedtreespeciescompositiononmanyplotsofourstudy,althoughanear-natural tree species compositionwas a precondition for the selec-tionofmostplots. Inourstudy,thehighest influenceofpastforestmanagement isexpected forP. abies in lowlandswhere thisspecieshasbeenfosteredbysilviculturalactivitiesatmanyplaces.However,aroundhalfofthe684plotswithoccurrenceofP. abiesare locatedabove1,000ma.s.l.wheretheabundanceofthisspeciesisnaturallyimportantonmanyforestsitesduetoharshclimate.Thus,carefulplotselectionandmanyhighaltitudinalplotsmighthavedampened thespurious impactof forestmanagementonthedata, theresults,andthestudyconclusions.
F IGURE 5 Speciessensitivitytolimitedsoilaeration.ResponsecurveswerederivedfromsinglevariableGLMswithcoverabundanceofmaturetreesastheresponsevariableandmeandepthofasoilwaterlevelduringthevegetationperiod(aproxyforsoiloxygenavailability)asapredictorbasedondatafrom1,075studyplotsacrossSwitzerland.Thedifferentspeciesabundancemaximareflectthedifferentmeanabundancethatspeciesreachedonthestudyplots(Table1).Fivespecies(Carpinus betulus, Ulmus glabra, Prunus avium, Quercus pubescens,andQuercus robur)arenotillustratedbecausetheirresponsecurvesrunhorizontallyatalevelofabout0%abundance
0.4
0.3
0.2
0.1
0.0
20015010050
Pred
icted
abun
danc
e
Depth of water level during vegetation period (cm)
Fag.syl
Abi.alb
Fra.exc
Pic.abi
Pin.syl
Que.pet
Ace.pse
| 9483WALTHERT And MEIER
5 | CONCLUSIONS
The edaphic predictors used in our study improved the statisticalmodel performance (i.e., themodel fit) for all studied tree species.Moreover, with two of our edaphic predictors, which quantifieddrought and soil oxygen shortage,wewere able to show that spe-ciesecologicalrequirementsassessedinlocalfieldstudiesandexperi-mentsaregenerallywell reflected in forestson larger scalesacrossSwitzerland.Thesefindingsunderlinetheecophysiologicalrelevanceoftheedaphicpredictorsusedinourstudy.
Ourthoroughlydata-drivenassessmentisexpectedtodeliveranobjective insight into species ecological niches. Itmay complementtheratherqualitativeknowledgeprovidedbythesystemofindicatorvalues.Ourresultsmaybeusedasbasicinformationforforestman-agementandhabitatprotection ingeneralandspecificallyforforestconversionunderachangingclimate.
ACKNOWLEDGMENTS
Wethank theWSL technical staff for samplingandanalyzing thesoilprofilesusedinthisstudy,and35botanistsforthecorrespondingveg-etationsurveys.WearegratefultoStefanZimmermann,MarcoWalser,andPeterJakobfortheircontributionstotheWSLforestsoildatabase,toDirkSchmatzforprovidingtheDaymetclimatedataandtoElisabethGrafPannatier,ManfredStähli,andMartinScherlerfortheirhelpinmod-elingthewaterbalance.WefurtherthanktheSwissFederalOfficefortheEnvironment(FOEN)forfinancingpartofthedataandMelissaDawesforeditingtheEnglish.TheevaluationswerepartlybasedondatafromtheSwissLong-termForestEcosystemResearchProgrammeLWF(www.lwf.ch),whichispartoftheUNECECo-operativeProgrammeonAssessmentandMonitoringofAirPollutionEffectsonForestsICP-Forests(www.icp-forests.net).WeareparticularlygratefultoAnneThimonierandElisabethGrafPannatier,whoprovideddatafrom16LWFplots.
CONFLICT OF INTEREST
Nonedeclared.
AUTHOR CONTRIBUTIONS
L.W.conceivedtheideas,providedthedata,andledthewriting,andE.S.M.analyzedthedataandassistedinwriting.
DATA ACCESSIBILITY
ThedataofthisstudyarearchivedinthesoildatabaseoftheSwissFederalInstituteforForest,SnowandLandscapeResearchWSLandare available on request. Contact person: LorenzWalthert ([email protected]).
ORCID
Lorenz Walthert http://orcid.org/0000-0002-1790-8563
REFERENCES
Anderegg,L.D.L.,&HilleRisLambers,J.(2016).Droughtstresslimitsthegeographicrangesoftwotreespeciesviadifferentphysiologicalmech-anisms.Global Change Biology,22,1029–1045.
Arend, M., Brem, A., Kuster, T. M., & Gunthardt-Goerg, M. S. (2013).SeasonalphotosyntheticresponsesofEuropeanoakstodroughtandelevateddaytimetemperature.Plant Biology,15,169–176.
Augusto,L.,Dupouey,J.L.,&Ranger,J.(2003).Effectsoftreespeciesonunderstoryvegetationandenvironmentalconditionsintemperatefor-ests.Annals of Forest Science,60,823–831.
Austin,M.P.,&Cunningham,R.B.(1981).Observationalanalysisofenvi-ronmentalgradients.Proc. Ecol. Soc. Austr.,11,109–119.
Austin,M.P.,&Smith,T.M.(1990).Anewmodelforthecontinuumcon-cept.InG.Grabherr,L.Mucina,M.B.Dale,&C.J.F.TerBraak(Eds.),Progress in theoretical vegetation science (pp. 35–47). Netherlands,Dordrecht:Springer.
Austin,M.P.,&VanNiel,K.P.(2011).Improvingspeciesdistributionmod-elsforclimatechangestudies:Variableselectionandscale.Journal of Biogeography,38,1–8.
Backes, K., & Leuschner, C. (2000). Leaf water relations of competitiveFagus sylvatica andQuercus petraea trees during 4years differing insoildrought.Canadian Journal of Forest Research- Revue Canadienne De Recherche Forestiere,30,335–346.
Beauregard,F.,&deBlois,S.(2014).Beyondaclimate-centricviewofplantdistribution:Edaphicvariablesaddvaluetodistributionmodels.PLoS ONE,9,e92642.
Borcard,D.,Legendre,P.,&Drapeau,P.(1992).Partiallingoutthespatialcomponentofecologicalvariation.Ecology,73,1045–1055.
Brassel, P., & Brändli, U.-B. (1999). Schweizerisches Landesforstinventar. Ergebnisse der Zweitaufnahme 1993-1995, Eidgenössische Forschungsanstalt für Wald und Schnee und Landschaft, Bundesamt für Umwelt und Wald und Landschaft.Bern:HauptVerlag.
Braun-Blanquet, J. (1964). Pflanzensoziologie. Grundzüge der Vegetationskunde. 3. Aufl.Wien,NewYork:Springer.
Cavin,L.,Mountford,E.P.,Peterken,G.F.,&Jump,A.S.(2013).Extremedroughtalterscompetitivedominancewithinandbetweentreespeciesinamixedforeststand.Functional Ecology,27,1424–1435.
Cools,N.,Vesterdal, L.,DeVos,B.,Vanguelova,E.,&Hansen,K. (2014).TreespeciesisthemajorfactorexplainingC:NratiosinEuropeanfor-estsoils.Forest Ecology and Management,311,3–16.
Coudun,C.,&Gegout,J.C. (2006).Thederivationofspeciesresponsecurves with Gaussian logistic regression is sensitive to samplingintensity and curve characteristics. Ecological Modelling, 199,164–175.
Coudun,C.,Gegout,J.C.,Piedallu,C.,&Rameau,J.C. (2006).Soilnutri-tional factors improvemodels of plant species distribution:An illus-trationwithAcer campestre (L.) inFrance.Journal of Biogeography,33,1750–1763.
Diekmann,M.,Michaelis,J.,&Pannek,A.(2015).Knowyourlimits–Theneedforbetterdataonspeciesresponsestosoilvariables.Basic and Applied Ecology,16,563–572.
Dreyer,E.(1994).Comparedsensitivityofseedlingsfrom3woodyspecies(Quercus roburL,Quercus rubraLandFagus silvaticaL)towater-loggingandassociatedroothypoxia:Effectsonwaterrelationsandphotosyn-thesis.Annales des sciences forestières,51,417–428.
Dubuis,A.,Giovanettina, S., Pellissier, L., Pottier, J.,Vittoz, P.,&Guisan,A. (2013). Improving thepredictionofplant speciesdistributionandcommunitycompositionbyaddingedaphictotopo-climaticvariables.Journal of Vegetation Science,24,593–606.
ESRI(2006).ArcGis 9.2.Redlands,CA:ESRI.Guisan,A., & Zimmermann, N. E. (2000). Predictive habitat distribution
modelsinecology.Ecological Modelling,135,147–186.Jobbagy,E.G.,&Jackson,R.B.(2000).Globalcontrolsofforestlineele-
vation inthenorthernandsouthernhemispheres.Global Ecology and Biogeography,9,253–268.
9484 | WALTHERT And MEIER
Köcher,P.,Horna,V.,&Leuschner,C.(2012).Environmentalcontrolofdailystemgrowthpatternsinfivetemperatebroad-leavedtreespecies.Tree Physiology,32,1021–1032.
Kolb,T. E.,&Matyssek, R. (2001). Limitations andperspectives aboutscaling ozone impacts in trees. Environmental Pollution, 115,373–392.
Kozlowski,T.T.(1986).Soilaerationandgrowthofforesttrees.Scandinavian Journal of Forest Research,1,113–123.
Kramer,P.J.,&Boyer,J.S. (1995).Water relations of plants and soils.SanDiego,CA:AcademicPress.
Kreuzwieser,J.,&Gessler,A. (2010).Globalclimatechangeandtreenu-trition:Influenceofwateravailability.Tree Physiology,30,1221–1234.
Kreuzwieser,J.,&Rennenberg,H.(2014).Molecularandphysiologicalre-sponsesoftreestowaterloggingstress.Plant, Cell & Environment,37,2245–2259.
Lendzion,J.,&Leuschner,C.(2008).GrowthofEuropeanbeech(Fagus syl-vaticaL.)saplings is limitedbyelevatedatmosphericvapourpressuredeficits.Forest Ecology and Management,256,648–655.
Lévesque,M.,Walthert, L., &Weber, P. (2016). Soil nutrients influencegrowth response of temperate tree species to drought. Journal of Ecology,104,377–387.
Lindner,M.,Fitzgerald,J.B.,Zimmermann,N.E.,Reyer,C.,Delzon,S.,vanderMaaten,E.,…Hanewinkel,M.(2014).ClimatechangeandEuropeanforests:Whatdoweknow,whataretheuncertainties,andwhatarethe implications for forest management? Journal of Environmental Management,146,69–83.
McCullagh, P., & Nelder, J. A. (1989).Generalized linear models, 2nd ed.London:ChapmanandHall.
Meier,E.S.,Kienast,F.,Pearman,P.B.,Svenning,J.C.,Thuiller,W.,Araujo,M.B.,…Zimmermann,N.E.(2010).Bioticandabioticvariablesshowlittleredundancyinexplainingtreespeciesdistributions.Ecography,33,1038–1048.
Mod,H.K.,Scherrer,D.,Luoto,M.,Guisan,A.,&Scheiner,S.(2016).Whatweuseisnotwhatweknow:Environmentalpredictorsinplantdistri-butionmodels.Journal of Vegetation Science,27,1308–1322.
Mood,A.M. (1971).Partitioningvariance inmultipleregressionanalysesasatoolfordevelopinglearningmodels.American Educational Research Journal,8,191–202.
MuellerDombois,D.,&Ellenberg,H.(1974).Aims and methods of vegeta-tion ecology.NewYork,NY:JohnWiley&Sons.
Niinemets, Ü., &Valladares, F. (2006). Tolerance to shade, drought, andwaterlogging of temperate northern hemisphere trees and shrubs.Ecological Monographs,76,521–547.
Nishimura,P.H.,&Laroque,C.P. (2011).Observedcontinentality in ra-dialgrowth-climaterelationships inatwelvesitenetwork inwesternLabrador,Canada.Dendrochronologia,29,17–23.
Normand,S.,Treier,U.A.,Randin,C.,Vittoz,P.,Guisan,A.,&Svenning,J.-C.(2009). Importanceof abiotic stress as a range-limitdeterminant forEuropeanplants:Insightsfromspeciesresponsestoclimaticgradients.Global Ecology and Biogeography,18,437–449.
Piedallu,C.,Gegout,J.C.,Lebourgeois,F.,&Seynave, I. (2016).Soilaer-ation,water deficit, nitrogen availability, acidity and temperature allcontribute to shaping tree species distribution in temperate forests.Journal of Vegetation Science,27,387–399.
Piedallu,C.,Gegout,J.C.,Perez,V.,&Lebourgeois,F.(2013).Soilwaterbalance performs better than climaticwater variables in tree spe-cies distribution modelling. Global Ecology and Biogeography, 22,470–482.
Pinto,P.E.,&Gegout,J.C. (2005).AssessingthenutritionalandclimaticresponseoftemperatetreespeciesintheVosgesMountains.Annals of Forest Science,62,761–770.
Prescott,C.E.(2002).Theinfluenceoftheforestcanopyonnutrientcy-cling.Tree Physiology,22,1193–1200.
R (2016).R: A language and environment for statistical computing.Vienna(Austria):RDevelopmentCoreTeam.
Rellstab,C.,Bühler,A.,Graf,R.,Folly,C.,&Gugerli,F. (2016).Usingjointmultivariateanalysesofleafmorphologyandmolecular-geneticmark-ers for taxon identification in three hybridizing Europeanwhite oakspecies(Quercusspp.).Annals of Forest Science,73,669–679.
Remund, J., Rihm, B., & Huguenin-Landl, B. (2014). Klimadaten für die Waldmodellierung für das 20. und 21. Jahrhundert. Schlussbericht des Projektes im Forschungsprogramm Wald und Klimawandel. Eidg. Forschungsanstalt fürWald, Schnee und LandschaftWSL. Retrievedfromhttps://doi.org/10.3929/ethz-a-010693673
Rigling,A.,Bigler,C.,Eilmann,B.,Feldmeyer-Christe,E.,Gimmi,U.,Ginzler,C.,…Dobbertin,M.(2013).DrivingfactorsofavegetationshiftfromScots pine to pubescent oak in dry Alpine forests. Global Change Biology,19,229–240.
Ryan,M.G.,&Yoder,B.J.(1997).Hydrauliclimitstotreeheightandtreegrowth.BioScience,47,235–242.
Scherrer,D.,Bader,M.K.F.,&Körner,C.(2011).Drought-sensitivityrank-ingofdeciduoustreespeciesbasedonthermalimagingofforestcano-pies.Agricultural and Forest Meteorology,151,1632–1640.
Schmull,M.,&Thomas,F.M.(2000).Morphologicalandphysiologicalre-actionsofyoungdeciduoustrees(Quercus roburL.,Q-petraea [Matt.]Liebl.,Fagus sylvaticaL.)towaterlogging.Plant and Soil,225,227–242.
Szymura, T. H., Szymura, M., & Maciol, A. (2014). Bioindication withEllenberg’sindicatorvalues:AcomparisonwithmeasuredparametersinCentralEuropeanoakforests.Ecological Indicators,46,495–503.
Thornton,P.E.,Running,S.W.,&White,M.A.(1997).Generatingsurfacesofdailymeteorologicalvariablesoverlargeregionsofcomplexterrain.Journal of Hydrology,190,214–251.
Thuiller, W. (2013). On the importance of edaphic variables to predictplantspeciesdistributions–Limitsandprospects.Journal of Vegetation Science,24,591–592.
Vesterdal,L.,Schmidt,I.K.,Callesen,I.,Nilsson,L.O.,&Gundersen,P.(2008).CarbonandnitrogeninforestfloorandmineralsoilundersixcommonEuropeantreespecies.Forest Ecology and Management,255,35–48.
Walthert, L., Graf Pannatier, E., & Meier, E. S. (2013). Shortage of nu-trients and excess of toxic elements in soils limit the distribution ofsoil-sensitive tree species in temperate forests. Forest Ecology and Management,297,94–107.
Weisberg,S.(1980).Applied linear regression.NewYork,NY:Wiley.Wohlgemuth,T.(2012).Vegetationdatabasesforthe21stcentury:Swiss
ForestVegetationDatabase. InJ.Dengler,J.Oldeland,F.Jansen,M.Chytry,J.Ewald,M.Finckh,F.Glöckler,G.Lopez-Gonzalez,R.K.Peet,&J.H.J.Schaminée(Eds.),Biodiversity & ecology(pp.340).Biodiversity,Evolution and Ecology of Plants (BEE), Biocenter Klein Flottbek andBotanicalGarden,UniversityofHamburg.
Zweifel,R.,Rigling,A.,&Dobbertin,M. (2009).Species-specificstomatalresponseoftreestodrought–alinktovegetationdynamics?Journal of Vegetation Science,20,442–454.
SUPPORTING INFORMATION
Additional Supporting Information may be found online in thesupportinginformationtabforthisarticle.
How to cite this article:WalthertL,MeierES.Treespeciesdistributionintemperateforestsismoreinfluencedbysoilthanbyclimate.Ecol Evol. 2017;7:9473–9484. https://doi.org/10.1002/ece3.3436