Ecology and Evolution. 2018;1–12. | 1www.ecolevol.org
Received:30November2017 | Revised:9March2018 | Accepted:15March2018DOI:10.1002/ece3.4057
O R I G I N A L R E S E A R C H
Applying ecological site concepts and state- and- transition models to a grazed riparian rangeland
Felix Ratcliff1 | James Bartolome1 | Luke Macaulay1 | Sheri Spiegal2 | Michael D. White3
ThisisanopenaccessarticleunderthetermsoftheCreativeCommonsAttributionLicense,whichpermitsuse,distributionandreproductioninanymedium,providedtheoriginalworkisproperlycited.©2018TheAuthors.Ecology and EvolutionpublishedbyJohnWiley&SonsLtd.
1DepartmentofEnvironmentalSciencePolicyandManagement,UniversityofCalifornia,Berkeley,Berkeley,California2USDA-ARS-JornadaExperimentalRange,LasCruces,NewMexico3TejonRanchConservancy,FrazierPark,California
CorrespondenceFelixRatcliff,DepartmentofEnvironmentalScience,Policy,&ManagementUCBerkeley,Berkeley,CA.Email:[email protected]
Funding informationTejonRanchConservancy
AbstractEcologicalsitesandstate-and-transitionmodelsareusefultoolsforgeneratingandtestinghypothesesaboutdriversofvegetationcomposition in rangelandsystems.Thesemodelshavebeenwidely implemented inupland rangelands,but compara-tively,littleattentionhasbeengiventodevelopingecologicalsiteconceptsforrange-landriparianareas,andadditionalenvironmentalcriteriamaybenecessarytoclassifyriparian ecological sites. Between 2013 and 2016, fifteen study reaches on fivecreekswerestudiedatTejonRanchinsouthernCalifornia.Datawerecollectedtodescribe the relationshipbetween riparianvegetationcomposition, environmentalvariables,andlivestockmanagement;andtoexploretheutilityofecologicalsitesandstate-and-transitionmodelsfordescribingriparianvegetationcommunitiesandforcreatinghypothesesaboutdriversofvegetationchange.Hierarchicalclusteranalysiswas used to classify the environmental and vegetation data (15 streamreaches×4years) intotwoecologicalsitesandeightcommunityphasesthatcom-prisedthreevegetationstates.Classificationandregressiontree(CART)analysiswasusedtodeterminethe influenceofabioticsitevariables,annualprecipitation,andcattleactivityonvegetationclusters.Channelslopeexplainedthegreatestamountofvariationinvegetationclusters;however,soiltexture,geology,watershedsize,andelevationwerealsoselectedasimportantpredictorsofvegetationcomposition.Theclassificationtreebuiltwiththislimitedsetofabioticpredictorvariablesexplained90% of the observed vegetation clusters. Cattle grazing and annual precipitationwerenotlinkedtoqualitativedifferencesinvegetation.Abioticvariablesexplainedalmost all of the observed riparian vegetation dynamics—and the divisions in theCARTanalysiscorrespondedroughlytotheecologicalsites—suggestingthatecologi-calsitesarewell-suitedforunderstandingandpredictingchangeinthishighlyvaria-blesystem.Thesefindingssupportcontinueddevelopmentofriparianecologicalsiteconcepts and state-and-transitionmodels to aiddecisionmaking for conservationandmanagementofrangelandriparianareas.
K E Y W O R D S
CARTanalysis,ecologicalsitedescriptions,grazingmanagement,hierarchicalclusteranalysis,riparianclassification
2 | RATCLIFF eT AL.
1 | INTRODUC TION
Riparian areas threading through upland rangelands boostlandscape-level biodiversity (Sabo etal., 2005), filter water (Tate,Atwill,Bartolome,&Nader,2006),andprovideothervaluableeco-system services (George, Jackson, Boyd,& Tate, 2011). They alsoprovideforageandwaterfor livestock,whichtendstocongregatetheseareas,potentiallydegradingriparianresources(Belsky,Matzke,&Uselman,1999;Kauffman&Krueger,1984).Accordingly,effortstoimprovetheoutcomesofriparianmanagementarecommon,butthehighlyvariableandsite-specificresponsesofrangelandriparianzonescancomplicatemanagers’abilitytomakereliablepredictionsabouttheeffectsofmanagement(Georgeetal.,2011).
Ecologicalsitedescriptionsandstate-and-transitionmodelsarecurrentlyregardedasusefulorganizingframeworksforunderstand-ingandpredictingthepatternsandprocessesonrangelands(Spiegaletal.,2016;Sayre,2017).Thesemodelshavebeenextensivelyde-velopedforuplandrangelandsintheUnitedStates,butonlyrecentlyhas attention been given to developing them for riparian systems(Stringham&Repp,2010).
Major determinants of riparian rangeland vegetation composi-tionincludefluvialprocessesandtheircontrolsonchannelgeomor-phology(McBride&Strahan,1984;Stella,Battles,McBride,&Orr,2010),depthtowatertableandsoilmoisturedynamics(Stringham,Krueger,&Thomas,2001), inundationfrequency (Sankey,Ralston,Grams, Schmidt, & Cagney, 2015), annual fluctuations in precipi-tation (Lunt,Jansen,Binns,&Kenny,2007),andflooddisturbanceregimes (Campbell&Green,1968).As a result of frequentdistur-bancesandspatialheterogeneity,vegetationwilllikelyneverreach“climax” stages (Campbell&Green, 1968), and biotic drivers suchas cattlegrazingmayhave limitedeffectsonvegetationcomposi-tion (Luntetal., 2007).Nevertheless, cattle tend to congregate inriparianareasandcanhaveexaggeratedeffectsonthesesystems(Kauffman&Krueger,1984),andmanagersneedmodelsthatcon-sider the role of abiotic disturbances, livestockmanagement, andsitepotential.
Rangelands inMediterranean-typeclimates,whicharepredict-ablymesicinthewinterandxericinthesummer,havedistinctflora,fauna and unique management systems, conservation challenges,andthreats(Bartolomeetal.,2014;Perevolotsky&Seligman,1998).RipariansystemsinMediterranean-typeregionshavehighinteran-nualandintra-annualweathervariationcoupledwitha“flashy”hy-drologyproducedduring the relatively shortwet season, creatingperiodicfluvialdisturbancesanddroughtwhichstructurebiologicalcommunities(Gasith&Resh,1999).Modelsthatconsidertheseabi-oticperturbationsmaybenecessarytodescribevegetationdynam-icsinMediterranean-typeripariansystems.
State-and-transitionmodelsareusuallyrepresentedbyboxandarrowdiagramsanddescriptivetextthatcatalogsalltheknownveg-etationstates(boxes)andtransitionsbetweenstates(arrows)foragivensite.Theyweredevelopedtomodelnonlinearvegetationdy-namicsinrangelandsystems(Westoby,Walker,&Noy-Meir,1989)and are a useful tool for communicating vegetation dynamics to
managers.Ecologicalsitesdescribedivisionsofthelandscapewithsimilarenvironmentalcharacteristicsthatsupportthesamerangeofstatesandtransitions (Spiegal,Larios,Bartolome,&Suding,2014).Given the variable nature of rangeland riparian sites, ecologicalsitedescriptionsandstate-and-transitionmodelsmaybe theopti-malframeworkforcataloguingandmakingpredictionsabouttheirecology(Stringham&Repp,2010)—butmoreinformationisneededabout how tobest classify ecological sites, states, andphases forrangelandriparianareas.
Upland sites are largely classified based on soils, climate, andlandscapeposition,whicharerelativelystableovertimescalesrele-vanttomanagement(Caudle,DiBenedetto,Karl,Sanchez,&Talbot,2013).Thesefactorsareprobablynotsufficienttodescribediffer-encesinrangelandripariansites,becauseripariansitesarealsoinflu-encedbydifferences in fluvialprocesses,channelgeomorphology,andhydrologiccyclesbetweensites(Stringham&Repp,2010).
Processesgoverningtemporalvariationwithinriparianecologi-calsitesdiffersomewhatfromthoseinuplandsaswell.Inadditiontoclimaticandmanagementdriversassociatedwithinterannualvaria-tioninuplands,fluvialprocessesandsoil–watercharacteristicsmaydrive temporal variation in vegetation composition (Stringham &Repp,2010;Stringhametal.,2001).Linkingcharacteristicsofchan-nelgeomorphology,soils,andhydrologicpropertiestodifferencesinriparianvegetationstatesisnecessarytohelppairriparianecologicalsitedescriptionswithstate-and-transitionmodels.
Given their value to conservation and management, it is im-portanttounderstandripariansystems inrangelandssothat theirmanagementcanbe improved.Thisstudyaddresses the followingresearchquestions:
1. Can ecological sites and state-and-transition models be usedto describe riparian vegetation assemblages and develop hy-potheses about their relationships to environmental and man-agement (i.e., cattle grazing) variables?
2. Inadditiontoparameterstypicallyusedinuplandecologicalsiteclassification,whatnewparametersareneededtoclassifyripar-ianecologicalsitesinaMediterranean-typesystem?
2 | MATERIAL S AND METHODS
2.1 | Study site
Tejon Ranch, located in southern California, contains 97,124 hec-taresofconservedlandsthatarejointlymanagedbytheTejonRanchCompany,TejonRanchConservancy,andtwograzinglessees.Cattlegrazing is themostwidespread landmanagement practice affect-ing riparian areas on the ranch. TheRanch encompasses areas ofCalifornia’s San Joaquin Valley, Sierra NevadaMountains,MojaveDesert,TehachapiMountains,andSouthCoastRanges.Thisstudyislimitedtomajorstreamswithwell-developedwoodyvegetation,intheSanJoaquinValleyportionoftheranch.Despitelarge-scalecon-versionofriparianforestsinCalifornia’sCentralValley,theseareas
| 3RATCLIFF eT AL.
provideawidearrayofecosystemservices(Vaghti&Greco,2007);duetoitsextentandmanagementhistory,TejonRanchprovidesan
ideallocationtostudyarelativelyintactnetworkofCentralValleyriparianforests.
Fivecreeksegmentswereselected forstudywithin theareaof interest:ChanacCreek (CH), El PasoCreek (EP), LowerTejonCreek (LT),TunisCreek (TU),andUpperTejonCreek (UT)—here-afterreferredtoas“creeksegments.”Withineachofthesecreeksegments, three locations were selected randomly within areaswithwoodyvegetationforatotalof15studyreaches—hereafterreferred toas “study reaches” (Figure1). In thewinterof2014–2015,onestudyreachoneachstreamsegmentwasrandomlycho-sentoreceiveacattleexclosure.Theexclosureswereinplacefortheremainderofthestudy.ReachesthatreceivedexclosureswereCH2,EP3,LT1,TU2,andUT3.
Although somewhat drier than the “true” Mediterraneanclimate (Aschmann,1984),thestudyarea is inaMediterranean-type region of California, with hot dry summers and cool wetwinters.Meanannualprecipitationis21cm.Eighty-ninepercentof this fallsbetweenNovemberandApril (http://ipm.ucanr.edu/WEATHER/wxactstnames.html). Mean maximum summer dailytemperatures are between 32 and 35°C, and mean minimumsummertemperaturesarebetween15and19°C.Meanmaximumwinter daily temperatures are between 15 and 21°C, andmeanminimumdailywintertemperaturesarebetween3and8°C.The4years encompassed by this study had below-average rainfall.Between 2012 and 2015, the average annual precipitation was
F IGURE 1 MapofstudyreachesonTejonRanch
Tejon Ranch
LT1 LT2LT3
CH1
CH2
CH3UT1
UT2UT3
EP1EP2
EP3
TU1TU2
TU3
LegendStudy reaches
Streams
Tejon Ranch
TABLE 1 Variablesusedintheecologicalsiteclusteranalysis
Variable Source Description
Elevation(m) Field Measuredinageographicalinformationsystem(GIS)frompointstakenusingaGPS
Slope(500m) GIS(10mspatialresolutionDigitalElevationModel[DEM])
Slopealonga500mportionofthecreekcenteredattheplotcenter(NRCSNationalCartographyandGeospatialCenter,2010)
Slope(thalweg) Field Slopeofthethalwegonthestudyreachtakenfromlongprofilefieldmeasurements
Sinuosity(500m) GIS DistancealongcreekdividedbyEuclidiandistancebetweenendpointsfor500mofcreek
Sinuosity(thalweg) Field DistancealongcreekdividedbyEuclidiandistancebetweenendpointsofthelongprofilemeasuredinstudyreach
WatershedSize GIS(10mDEM) Areaofwatershedcontributingtostreamatstudyreach(NRCSNationalCartographyandGeospatialCenter2010)
Geology GIS Mappedgeologyatstudyreach.(Dibblee,2005,2008a,b)
DominantUpstreamGeology GIS Mostcommongeologymappedalongstreambetweenstudyreachandheadwaters.(Dibblee,2005,2008a,b)
Width:DepthRatio Field Widthofstreamchanneldividedbydepthofchannel.Calculatedfromchannelcross-sections
EntrenchmentRatio Field Widthofflood-proneareadividedbywidthofchannel.Calculatedfromchannelcross-sections
GreenlineHeightAboveThalweg Field Averageheightofgreenlineabovethalweginthestudyreach.
Sand(%) Field Percentsandincompositesoilsamplealonggreenline
Silt(%) Field Percentsiltincompositesoilsamplealonggreenline
Clay(%) Field Percentclayincompositesoilsamplealonggreenline
4 | RATCLIFF eT AL.
only15.7cm.The2015–2016rain-yearhadapproximatelyaver-agerainfall(20cm).
2.2 | Sampling abiotic site factors
Fourteenvariableswereusedtoclassifythe15studyreachesintoeco-logicalsites(Table1).Theseincluderemotelysensedvaluesmeasuredinageographicalinformationsystem,suchaselevation,slope,sinuos-ity,watershed size, andgeology (Table1).Measurementsof streamgeomorphologyweremadeinthefieldusingatotalstation.Soilsam-pleswerecollectedinthefieldin2013andanalyzedforsoiltextureusingthehygrometermethodattheUCDavisAnalyticalLaboratory.
2.3 | Sampling vegetation
VegetationwassampledannuallyateachofthestudyreachesinlateMayandearlyJuneof2013through2016.A“greenline”transectfol-lowedthetoeofthecreekbankandsampledvegetationgrowingnearthewater’sedge.Winward (2000) recommendssamplinggreenlinevegetationatthetopofbank;however,wesampledthegreenlineatthetoeofthebankbecausetheherbaceousvegetationatthetopofbankwastypicallycomposedofthesameannualgrassspeciesthatdominate theadjacentuplands. Inorder to sample theherbaceousspecies compositionmost influencedby the stream,weneeded tosampleinthewettersoilsfoundatthetoeofthebank.Shrubsandtreeswererecordedinthesamplingtransectsregardlessofposition.
Greenlinevegetationcompositionwasmeasuredalong50mofthe15studyreachesinthreedifferentstrata:herbaceous,shrub,andtree.Thestartingpointforeachtransectwastherandomlyassignedcenterpointofthestudyreach.Herbaceousvegetationwasmeasuredwithaline-pointintercepttransect.Eachhalf-meteralongthetransecttape,thefirstplanthitwithinonemeterabovethegroundwasrecorded.Aline-intercepttransectwasusedtorecordthelineardistanceofshrubsand treesoverhanging thegreenline transect.Anyplant (regardlessofspecies)overhangingthetapebetweenoneand3mofheightwasrecordedinthe“shrub”category,andanyplantoverhangingthetapeabove3minheightwasrecordedinthe“tree”category.
2.4 | Statistical methods
Our analytical approach proceeded as follows: (1) cluster analysiswasusedtoclassifystreamreachesintoecologicalsiteswithrespecttotheirabioticsitefactorsand(2)todefinemeaningfulvegetationcommunityassemblages (vegetationclusters); (3) IndicatorSpeciesAnalysiswasusedtodescribethevegetationclusters;(4)thepres-enceofstateswereevaluatedasaggregationsofclusters;(5)CARTanalysiswasusedtoidentifytheinfluenceofabioticsitefactorsonvegetationclusterdifferentiation.
2.5 | Ecological site cluster analysis
Hierarchical cluster analysis can be used to classify ecologicalsitesbasedongroupingsofkeyabioticenvironmentalvariables in
rangelanduplands(Spiegaletal.,2014).Similarly,ithasbeenusedtoclassifystreamreachesfromgeomorphicandhydrologicmeasure-mentsofstreamchannelsand isespeciallyuseful ifappliedwithinadistinctphysiographicunitwhereitcanyieldobjectiveclassifica-tions (Kondolf,Montgomery, Piegay, & Schmitt, 2003). A suite ofindicators used in stream classification and those used in uplandecologicalsiteclassificationwerecombinedinaclusteranalysistocreatetheriparianecologicalsiteclassification(Table1).
TheecologicalsiteclusteranalysiswasperformedusingGower’sdistance,whichcalculatessimilarityforeachvariableinthematrixsep-arately(usingamethodaccordingtothevariabletype)andisthereforeabletoanalyzebothcontinuousandcategoricalvariablestogether.Thefinaldistancemetricisanaverageofthepartialsimilarities(Borcard,Gillet,&Legendre,2011).AnalysiswasperformedinRusingthepack-ages “vegan” and “cluster” (Maechler, Rousseeuw, Struyf,Hubert,&Hornik,2015;Oksanenetal.,2015;RCoreTeam2016).Theclusterdendrogram was pruned using the Mantel test, which compares amatrixofclusterassignmentstotheoriginaldistancematrixusedtocreatetheclusterdendrogram.Thistestisrepeatedforeverypossiblenumberofclusters,andthenumberwiththehighestMantelcorrela-tionisconsideredtheoptimalnumberofclusters(Borcardetal.,2011).
2.6 | Vegetation cluster analysis
Aclusteranalysiswasperformedonthegreenlinevegetationcoverdata to investigate patterns of riparian plant community structurewithinthe15studyreachesover4years.The60uniqueReach×Yearcombinationswereclusteredbasedonabsolutecoverofallliveplantsalongtransects.Proportionalcoverdatafromtheshrubandtreelay-ersweregenerallymuchhigherthandatafromtheherbaceouslayer;therefore, herbaceous layer datawere square-root transformed sothattheshrubandtreelayersdidnotoverlyinfluencetheclusteras-signments(McCune,Grace,&Urban,2002).ShrubandTreecoverwasnottransformed.Similarly,allspeciesoccurringon<2Reach×Yearswere removed from the analysis so that very rare species did notdisproportionately influencetheanalysis.Allspecies×canopyclasscombinationswere treated as unique species. The cluster analysiswasperformedusingBray–Curtisdistance,whichcalculatessimilar-itybasedonspeciesfoundtobepresentonstudyreachesratherthanmutualabsences(Zuur,Ieno,&Smith,2007).
Twomethodswere utilized to prune the cluster dendrogram totheoptimalnumberof clusters.First, aMantel correlation testwasused,asdescribedabove.Second,anIndicatorSpeciesAnalysiswasperformed,andthenumberofgroupswhichcontainedthemostsig-nificant indicator specieswas selected (Dufrene& Legendre, 1997;McCuneetal.,2002).AllstatisticalanalyseswereperformedinRusingpackages: vegan, cluster, and indicspecies (DeCaceres&Legendre,2009;Maechleretal.,2015;Oksanenetal.,2015;RCoreTeam2016).
2.7 | Indicator species analysis
In addition to showing the optimal location to prune the clusterdendrogram,the“significant”indicatorspeciesshowwhichspecies
| 5RATCLIFF eT AL.
bestcharacterizeeachcluster. Indicatorspeciesarethosethatarecommonwithinstudyreachesofonecluster,andrelativelyscarceinstudyreachesofotherclusters(Dufrene&Legendre,1997).Basedon these criteria, species aregivenan indicatorvalue (0–1), andarandomizationtestisperformedtodeterminethestatisticalsignifi-canceoftheindicatorvalue.“Significant”indicatorspeciesarethosewith<5%probabilityofhavingnodifferencebetweengroups.
2.8 | Identifying states, phases, and transitions
The USDA Natural Resources Conservation Service (NRCS) im-plementation of state-and-transition models and ecological sitedescriptions is largely a “top-down” process where elements ofstate-and-transition models are drawn and populated by expertopinionandafterwardvalidatedwithdata (Jackson,Bartolome,&Allen-Diaz,2002).Incontrast,inthisstudy,plantspeciesdataareag-gregatedtobuildvegetationstatesfromthe“ground-up”withfewerpreconceptions about what constitutes a “state.” The NRCS alsodifferentiates betweenminor, easily reversible changes in vegeta-tionlabeled“phase-shifts”andthemoreresilient“states”theyoccurin (Bestelmeyer etal., 2003; Stringham,Krueger,&Shaver, 2003).These distinctions formalize some general aspects of the originalstate-and-transitionapproachandprovideusefulcategoriesthatcanbethebasisoftestablehypotheses.
Inthisstudy,weperformedclusteranalysistodefinemeaningfulvegetationassemblages.Manyoftheseclustershadsimilarvegeta-tionstructureandfunctionalgroupcomposition—thushavingsimi-larimplicationsformanagement—andtransitionsbetweensomeoftheseclusterswouldlikelyoccurwithoutthresholddynamics.Asaresult, theoptimalnumberof clusters from thevegetationclusteranalysiswasconsideredvegetation“phases,”not“states.”Themoregeneral “states” were defined by considering: potential drivers ofspatial and temporal variation (e.g., irreversible geomorphologicalchanges),differencesinBray–Curtisdistancebetweentheclusters,and ecological characteristics of the dominant and indicator spe-ciesofeachcluster (e.g.,wetlandvs.uplandplants).The resultingstatesarestillbasedontheoriginalvegetationclusterdendrogram,butrepresentadeeper“cut”ofthedendrogramwithfewerterminalnodes.
Thevegetationclusteranalysiswasperformedondatafromall15studyreaches,andtheresultingstatesandphasesweresubse-quently divided into the two ecological sites. This procedurewaschosenbecause(1)itallowedevaluationofhowwelltheecologicalsitescorrespondedtoobserveddifferencesinvegetationdynamics,and(2)althoughstudyreachesarerepresentedbydiscreetecolog-ical sites, they represent a gradientof site characteristics and arethereforeexpectedtosharesomevegetationstates.Combiningdatafromallstudyplotsshowedwhichstatesareuniquetoeachecolog-icalsite,andwhicharesharedbetweenthem.
Inourscheme,a“temporaltransition”occurswhenthestateatastudyreachmovesinspeciesclusterspacebetweenyears(sensuSpiegal etal., 2014). “Spatial transitions” are evident in cases inwhichdifferentvegetationclustersoccur indifferentareaswithin
the same ecological site and are differentiated by spatial—insteadof inherently temporal—processes (also seeBestelmeyer,Goolsby,&Archer,2011).
2.9 | Classification Tree (CART)
Aclassificationtreewasbuilttodeterminewhichenvironmentalfac-torsbestpredicttheobservedvegetationstatesandtoinformourecological site classification approach. The response variable (thedata tobepartitioned)was the clusters from thevegetation clus-teranalysis,andtheindependentvariablesweretheenvironmentalvariablesusedintheecologicalsiteclusteranalysis,annualprecipi-tation,andgrazingtreatments(exclosures).Aclassificationtreeusestop-down recursivebinary splitting topartition the responsedataintoatreethatoptimizestheclassificationofresponsevariablesateachnodewith respect toeachof thepredictorvariables (James,Daniela,Trevor,&Robert,2013).
The classification treewas built using the “tree” package in R(Ripley, 2016). The tree was pruned using the function “cv.tree,”whichdeterminestheoptimalnumberofterminalnodesbyminimiz-ingthedevianceinaK-foldcross-validation(Ripley,2016).PruningtheCARTtreetoseventerminalnodesresultedinthelowestdevi-anceintheCARTanalysis.Thisresultedinonlysixofthefourteenabioticfactorsbeing includedintheconstructionoftheclassifica-tiontree(Table2).
3 | RESULTS
3.1 | Ecological sites
The Mantel correlation test showed that the optimal number ofclusterswas 2 (r=.644), representing two ecological sites: LowerTejonCreekandallotherstudyreaches(Figure2).Thervalueforthenexthighestcorrelation(forfiveclusters)wassubstantiallyloweratr=.572.
Thesetwoecologicalsitesdifferinseveralregards.EcologicalSite 1 is morewidespread in the study area and as a result ismorevariable.ReachesinEcologicalSite1(allstudyreachesex-ceptthoseonLowerTejonCreek)havehigherelevations,higherchannel slopes, smaller watershed sizes, lower entrenchmentratios,moresiltand lesssand in thesoil, andmorediversege-ologies andupstreamgeologies than reaches inEcological Site2 (thoseonLowerTejonCreek).Thevariablesthatdonotsub-stantiallydifferbetween the twoecological sitesaresinuosity,width:depthratio,greenlineheightabovethalweg,andpercentclayinsoil(Table2).
3.2 | Vegetation states
The vegetation cluster analysis showed that Reach×Years gener-allyclusteredmostcloselywiththesamereachinotheryears.TheMantel correlation test pruned the resulting dendrogram to 10clusters(r=.663).However,eightclustershadthemostsignificant
6 | RATCLIFF eT AL.
indicator species p-values andwere therefore selected by indica-torspeciesanalysis.AstheMantelcorrelationcoefficientwasveryclosebetweeneightandtenclusters,(r=.645andr=.663respec-tively), eight clusters were selected to represent the vegetationgroups(Figure3).
Each of the eight clusters has statistically significant indicatorspecies.Allclusters includeperennialwoodyspeciesas indicators,andallclustersexceptClusters1and2includeherbaceousspeciesas significant indicator species (Figure4). Indicator species are al-waysmostabundantintheclustertheyareassignedto;however,in
Ecological site 1 Ecological site 2Selected by CART model?
Elevation(m) 444 265 Yes
Slope(500m) 3.1% 1.7% Yes
Thalwegslope 2.8% 1.0% No
Sinuosity(500m) 1.10 1.15 No
SinuosityofThalweg 1.20 1.14 No
Watershedsize(m2) 9.68×107 2.82×108 Yes
Width:Depthratio 8.49 9.97 No
Entrenchmentratio 3.39 5.86 No
GreenlineheightaboveThalweg(m)
0.213 0.223 No
Sand(%) 79.5 87.0 Yes
Silt(%) 14.83 7.33 Yes
Clay(%) 5.67 5.67 No
Geology Qt,gn,hdq1 Qa Yes
Dominantupstreamgeology
Qt,gn,hdq Qa No
Geology abbreviations indicate the following geology map units (Dibblee, 2005, 2008a,b):gn=Gneissic Rocks; hdq=Mafic Intrusive Rock; Qa=Quaternary Alluvium; Qt=QuaternaryTerraceDeposits.
TABLE 2 Meanvaluesperecologicalsiteforeachoftheenvironmentalattributesusedintheecologicalsiteclusteranalysis.Geologyvalueslistallthegeologymappingunitsineachecologicalsite
F IGURE 2 Clusterdendrogramshowingtheresultsoftheecologicalsiteclusteranalysisbasedonabioticvariables.Themorecloselytheirbranchesarerelatedinthedendrogram,themoresimilartheirenvironmentalattributesare.Thelettersinthestudyreachnamesrepresentthecreekstheyareon.CH=ChanacCreek,EP=ElPasoCreek,LT=LowerTejonCreek,TU=TunisCreek,UT=UpperTejonCreek.ThedottedlineshowstheoptimallocationtoprunethedendrogramEcological Site 2Ecological Site 1
Ecological site clusters -- Gower distace
| 7RATCLIFF eT AL.
thisanalysis,theygenerallyoccurinotherclustersaswell,sotheirmerepresenceisnotdiagnosticofclustermembership.Justfiveofthe41 indicator speciesoccurred inonlyonecluster, and fiveoc-curred in all eight clusters. Thirty-five of the 41 indicator specieswereinthetopfivespecies(bycover)fortheircanopylayerintheclustertheybelongedto.
Per methods described previously, three vegetation “states”weredefinedamongtheeightvegetationclusters(Figure3).States1and2occurexclusivelyinEcologicalSite1,whileState3occursalmostexclusivelyinEcologicalSite2,buthasalimiteddistributiononEcologicalSite1(Figure3).Thesethreestatesrepresentadeeper“cut” of the dendrogram and also have a highMantel correlationvalue(r=.61).Thethreestatesare:
3.2.1 | Vegetation state 1
This state comprises four of the eight vegetation clusters(Clusters1,5,6,and8)thatarecloselybranchedontheclusterdendrogramandisonlypresentinEcologicalSite1(Figure3).Thereaches in this state (CH1, EP1, EP2, EP3, TU1, TU2, TU3, andUT1 in all years) generally hadmultitiered canopies of ripariantrees: Salix laevigata, Populus fremontii,Quercus lobata, and thevine Vitis californica; and these clusters have many hydrophilicindicator species (Figure4). Their close linkagedistances in theclusteranalysisaswellastheecologicalsimilaritiesoftheirindica-torspeciessuggestthatchangesbetweentheseplantcommunitytypesmayhappenfrequentlyandwithoutmajoroutsideforcing.For that reason, thefourclustersare included in thestate-and-transitiondiagramascommunityphaseswithinVegetationState1(Figure4).
3.2.2 | Vegetation state 2
Althoughmostof the study reachesclassifiedasEcologicalSite1areclusteredinarelativelycohesiveareaofthevegetationclusterdendrogram,onecluster(Cluster3)isisolatedfromtherestoftheEcologicalSite1clusters(Figure3).UnlikeVegetationState1,veg-etation inCluster3 is characterizedbyuplandannualgrassesandforbs intheherbaceouslayer,andlacksanywoodyplant indicatorspeciesexceptthenon-nativeshrubNicotiana glauca.Noneof theindicator species in Cluster 3 is considered hydrophilic (Figure4).Thestudyreachesthatmakeupthisclusterarethepredominantlydryreaches:UT2(allyears),UT3(allyears),andCH2(in2015and2016only).ThegenerallydryconditionsandephemeralstreamflowonUT2andUT3makeitunlikelythatthesereacheswillshifttoacommunity dominated by woody plants and hydrophilic speciescharacteristic of the other vegetation clusters in Ecological Site 1withoutamajorweathereventandasubsequentchangeingeomor-phologyandhydrology.For that reason,Cluster3 is includedasauniquevegetationstate(State2)inthestate-and-transitiondiagram(Figure4).
3.2.3 | Vegetation state 3
InEcologicalSite2,therearethreecloselyrelatedvegetationclus-ters (clusters2,4,and7).All threeshareacommonbranchoftheclusterdendrogram (Figure3) and share similar riparian shrubandtree species.Theclustershave relativelyhighcoverofSalix good-ingii,Populus fremontii,andBaccharis salicifolia,relatively lowcoverof Salix laevigata, and no Quercus lobata cover. Given the simi-larities in perennial riparian vegetation and their proximityon the
F IGURE 3 Clusterdendrogramshowingtheresultsofthevegetationclusteranalysis.TheunitsbeingclusteredarealltheReach×Years.Themorecloselyrelatedbranchesinthedendrogramhavemoresimilarvegetation.ThelettersintheReach×Yearnamesrepresentthecreekstheyoccuron.ReachnamesinblackareinEcologicalSite1;namesingrayareinEcologicalSite2.Thesolidanddottedlinesshowwheretotrimthedendrogramtoproduceeightcommunityphaseclustersandthreevegetationstates,respectively
Cluster 1 Cluster 6 Cluster 8 Cluster 2Cluster 7 Cluster 4 Cluster 3Cluster 5
State 1 State 3 State 2
8 | RATCLIFF eT AL.
clusterdendrogram,thesethreevegetationclustersareallconsid-eredphasesinVegetationState3(Figure4).
3.3 | Transitions and phase shifts
Spatialvariationwithineachof theecologicalsiteswasmorepro-nouncedthantemporalchangeoverthestudyperiod.Intotal,sevencommunity phases (i.e., the vegetation clusters) comprising threevegetationstateswereobservedacrossthereachesinEcologicalSite
1,andthreecommunityphaseswereobservedacrossthereachesinEcologicalSite2(Figure4).Ofallthepotential“spatialtransitions,”compelling evidenceonly exists for the causeof one transition inEcologicalSite1betweenVegetationState2andVegetationState1(T2,Figure4).InEcologicalSite1,oneminor“temporal”phaseshiftandonemoresignificant“temporal”transitionwerealsoobservedoverthe4yearsofthestudy;andonlyonephaseshiftwasobservedonreachesinEcologicalSite2.Asummaryofthesetransitionsandphaseshiftsisbelow:
F IGURE 4 State-and-transitiondiagramfortheriparianstudyreaches.ThetopdiagramshowsthestatesandphasesoccurringonstudyreachesinEcologicalSite1,andthebottomdiagramshowsstateandphasesoccurringonreachesinEcologicalSite2.Specieslistedineachphasearethesignificantindicatorspeciesforthatphase,listedbydescendingorderofindicatorvalue.Solidarrowsindicate“temporaltransitions”andphaseshifts,thedottedarrowshowstheonly“spatialtransition”withaplausibledriver.Wetlandcodesareprovidedinparenthesesaftereachspeciesname(Lichvar,Banks,Kirchner,&Melvin,2016).An*indicatesthatthespeciesisnotincludedin“TheNationalWetlandPlantList”(Lichvaretal.,2016).Thewetlandstatusofspecieswithone*isinferredfromcongenersonthelist.Specieswithtwo**donothavecongenersonthelist,andtheirwetlandstatusishypothesizedfromauthors’fieldobservations.MoreinformationonplantspeciesisincludedinTableS1.DescriptionsofthestatesandtransitionsareinthetextoftheResultssection
| 9RATCLIFF eT AL.
3.3.1 | Transition from vegetation state 2 to vegetation state 1 (spatially- observed) (T2-Ecological Site 1)
ThistransitionoccursinEcologicalSite1whenVegetationState2(VegetationCluster3—characterizedbydry stream reachesdomi-nated by upland annual grasses) changes to Vegetation State 1(Vegetation Cluster 6—characterized by perennially wet streamreachesdominatedbyhydrophilicplantspecies)(Figure4).AlthoughthesetwostatesbothoccuronUpperTejonCreek,theyaresepa-ratedbyalargeheadcutandhavedrasticallydifferentchannelmor-phologiesandstreamflows.Abovetheheadcut,studyreachesUT2and UT3 are multichannel reaches with short distances betweenthalwegandhistorical floodplains (2.0and1.5mrespectively).Onthese reaches,water flowsephemerallyanduplandplantsare thedominantvegetation.Below theheadcut,UT1 isa singlechannelreach with a much larger distance between thalweg and historicfloodplain(7.2m).Onthisreach,waterflowsyear-roundandvegeta-tionisamixofwetlandplants.
3.3.2 | Transition from vegetation state 3 to vegetation state 2 (temporally- observed) (T1-Ecological Site 1)
InEcologicalSite1,thestudyreachCH2changedfromVegetationState3 (VegetationCluster 2—characterized byPlatanus racemosa and Baccharis salicifolia) in 2013 and 2014 to the upland annual-dominatedVegetation State 2 (VegetationCluster 3) in 2015 and2016 (Figure4).Althoughthis transitionoccurredonaplotwitha
cattleexclosure,itisunlikelythatthechangesincattleactivitypre-cipitatedthischange.Thetransitionwascharacterizedbyadie-offofestablishedripariantreesandwasmorelikelytheresultof4yearsofbelow-averageprecipitation.Duringtheseyears,Populus fremontii absolutecoverinthetreecanopydecreasedfrom40%in2013and2014to5%in2015and0%in2016;Salix laevigatacoverinthetreecanopydecreasedfrom35%in2013to13%in2015and0%in2016(Figure5, Figure S1). Reversing this transitionmay require severalyearsofwetconditionstoreestablishthesetreespecies.
3.3.3 | Phase shift from vegetation cluster 1 to 6 (temporally- observed) (PS1-Ecological Site 1)
InEcologicalSite1,thestudyreachEP3changedfromVegetationCluster1toVegetationCluster6between2015and2016sampling.ThisrepresentedaphaseshiftfromacommunitydominatedbyVitis californicaandSalix laevigataintheherbaceouslayertoonecharac-terizedbya suiteofherbaceoushydrophilicplants (Figure4).Thisphaseshiftfollowedanunusualsummerfloodin2015thatclearedoutsomeofthewoodyplantunderstory.
3.3.4 | Phase shift from vegetation cluster 7 to 2 (temporally- observed) (PS1-Ecological Site 2)
The only phase shift observed in Ecological Site 2waswhen thestudyreachLT2changedfromVegetationCluster7 toVegetationCluster2between2013and2014sampling.ThisrepresentedashiftfromacommunitycharacterizedbySalix goodingii,Populus fremontii,andasuiteofherbaceoushydrophytestoacommunitycharacter-izedbyhighcoverofBaccharis salicifoliaintheherbaceouscanopy.TheshiftoccurredafterherbaceoushydrophyteandSalix goodingii coverdecreasedin2014,possiblyalsotheresultofbelow-averageprecipitation(Figure4).
3.4 | Results of CART analysis
Therootsplit in theclassificationtreewaschannelslope (500m),indicatingthatitexplainedthemostvariationinvegetationphases.After that, a combinationof soil texture, geology,watershed size,andelevationwerethefactorschosentofurtherpartitiontheclus-terassignments.Thereach-scalestreamgeomorphologicalmeasure-ments,cattleexclosures,andannualprecipitationwerenotincludedintheprunedclassificationtree,indicatingthattheydidnotconsist-entlypredictthedifferentvegetationclusters(Figure6).Overall,theprunedCARTmodelcorrectlyclassified90%oftheReach×Years,withonlysixReach×Yearsmisclassified.
4 | DISCUSSION
The distribution of the vegetation states and phases was largelyexplained by the two ecological sites (Figures3 and 4). Similarly,phasesandstatesfromeachoftheecologicalsitesoccurredlargely
F IGURE 5 PhotosofstudyreachCH2.Notethedie-backintreecanopybetween2013and2016.VegetationonthereachwasclassifiedasVegetationState3in2013and2014,butchangedtoVegetationState2in2015and2016.Thereisnophotofor2015,buttreeandshrubthinningwasobserved
2013
2014
2016
10 | RATCLIFF eT AL.
onseparatebranchesoftheclassificationtree(Figure6),suggestingthatecologicalsitecharacteristicscorrespondedcloselywithdiffer-ences in vegetation.This couplingof ecological siteswithvegeta-tionphasesandstatesissignificantasitshowsthattheecologicalsiteclassification(basedpurelyonabioticvariables)canbeusedtoexplaintheoccurrenceofalmostallofthevegetationphasesonthestudyreachesandshowsthateachecologicalsitesupporteduniquevegetationstates.
The 15 study reaches are on the boundary of two floristic re-gionsinCalifornia:(1)theSanJoaquinValleySubregion,and(2)theTehachapiMountainAreaSubregion(Baldwin,2012).TheyarealsoontheboundaryoftwoMajorLandResourceAreas(MLRAs).Thehigherelevation reaches in Ecological Site 1 are closer to the TehachapiSubregion and fall in the Sierra Nevada Foothills MLRA, and thereaches in Ecological Site 2 fall squarely in the San JoaquinValleySubregionandareclosertotheSacramentoandSanJoaquinValleyMLRA.Whileitisnotsurprisingthatthetwoecologicalsites—whichdescribeanelevationgradientontheboundaryoftheregions—havedistinctvegetationcommunities,itisvalidatingthattheyalsodetectdifferencesinvegetationpredictedbythesebiogeographicregions.
Theclassificationtreehighlightedwhichenvironmentalfactorscorrespondedtodifferencesseeninthevegetationcommunitiesandgaveagoodindicationofhowbesttoadaptexistingecolog-ical site concepts to create meaningful predictions for riparianrangelandsinthisregion.Itshowedtheimportanceoftheunder-lyingsitevariables:channelslope,elevation,geology,watershedsize and soil texture; suggesting that in this system—in additiontouplandecologicalsitecriteria—channelslopeandcontributorywatershed size should be considered when classifying riparianecologicalsites.
Surprisingly,with the exception of channel slope, stream geo-morphologicalfactorswerenotimportantintheclassificationtree.
This implies that none of these geomorphological variables con-sistently predicted differences in the vegetation phases. Streamcross-sectionalprofilesdifferedatlargespatialscales,asseeninthedifferencesbetweenecologicalsites(Table2),buttheyalsovariedatrelativelysmallspatialscalesthroughoutthestudyarea(e.g.,be-tweenstudyreachesonacreeksegment).Channelgeomorphologywasonlymeasuredonceateachreachduringthestudy;however,itwasconsideredrelativelystableoverthestudyperiodbecauseofthebelow-averagerainfall.
Cattleexclosureandprecipitationwerealsonotsignificantvari-ables in theCARTanalysis. Thismakes sense given that (1) cattleexclosureswereonlyinplacefortwoyears;(2)rainfallwasnothighlyvariableoverthestudyperiod;and(3)clusterindicatorspeciescon-tainedmanyperennialwoody species. The apparent lack of influ-encefromcattlegrazingraisesimportantquestionsformanagementofthissystem,including:
1. Does cattle activity affect vegetation states over longerperiodsof time? Ifso,what levelsofsustainedcattleactivityresult in qualitative changes to vegetation, and what arethe primary mechanisms of change (e.g., hindering woodyplant recruitment)? Are these changes contingent on eco-logical site?
2. Arecertainstatesorphasesmoresensitivetocattleactivity?Inparticular,are reacheswithmoreherbaceous,perennialhydro-phytes (e.g.,Cluster6 inEcologicalSite1)more likelytobeaf-fectedbycattle?
3. Whatopportunitiesexistforenhancingormaintainingriparianvegetation?Arethereinteractionsbetweengrazingandrainfallorgrazingandacornmastyears that shouldbe takenadvan-tageoforavoided?DoesgrazingmanagementinUpperTejonCreek affect the rate at which the head cut in that stream
F IGURE 6 Resultsfromaclassificationtreewithvegetationclustersasthecategoricalresponsevariable,andtheabioticEcologicalSitevariables,totalannualprecipitation,andcattleexclosuresasthefactorsusedtosplitthedata.Thesplitsfartherupinthetreeexplainmoreoftheoverallvariationinvegetation
Slope_500 m < 2.3%
Geology: Qa
Elevation < 272.5 m
Sand < 78.5%
Watershed < 8.8e + 07 m2 Silt < 12%
Elevation < 378 m
7 4
3
8 4
6 15
| 11RATCLIFF eT AL.
segment moves upstream—and thus the vegetation state?Could management strategies—such as seasonal grazing re-gimes,bankstabilizationusingrestorationplanting,ormoder-ating peak stream discharge using an existing dam—slow orstop the movement of the head cut and prevent vegetationfromtransitioningfromState2toState1?
These questions can be formulated as formal hypotheses andtestedthroughlonger-termmonitoringofexclosuresorriparianpas-tureswithprescribedstockingrates.Theecologicalsitesandvegeta-tionstates/phases identified inthisstudyprovideecologicalcontextthatcanguidemanagers’selectionofstudylocations,treatments,andmonitoring methods to efficiently answer these questions. Furtherinvestigation in this areawill result in better descriptions of states,driversoftransitions,andecologicalsiteboundaries.Futureresearchshouldalsoinvestigatethedistributionandextentoftheseecologicalsitesandtheirassociatedstate-and-transitiondynamicsinthesouth-ernSanJoaquinValleyandtheSierraNevadaFoothillsMLRA.
5 | CONCLUSIONS
By including riparian-specific criteria, ecological site classificationscanbebuiltforripariansystems.OnTejonRanch,riparianecologicalsitedescriptionsandstate-and-transitionmodelsprovidedaunifiedframeworklinkingabioticandmanagementfactorstovegetationdy-namics.Thesemodelswereableto incorporateandorganizehighlyvariableripariansitefactorsandvegetationassemblages.Bycatalog-ingknownphases,states,andtransitionsoneachecologicalsite,thesemodelscreatedanorganizedapproachtounderstandingthecomplexandsite-specificresponsesofrangelandriparianareas.Theyprovidedaframeworkforpredictingvegetationstatesandtransitions,andforgeneratingandtestinghypotheseslinkingweather,management,andsitecharacteristicstovegetationchangesovertimeandspace.
ACKNOWLEDG MENTS
This work was supported by the Tejon Ranch Conservancy. Wealso thank Michele Hammond, Peter Hopkinson, and the UCBerkeleyRangeEcologyLab for their feedbackandcontributionsto fieldwork. Publication made possible in part by support fromtheBerkeleyResearchImpactInitiative(BRII)sponsoredbytheUCBerkeleyLibrary.
CONFLIC T OF INTERE S T
Nonedeclared.
AUTHOR CONTRIBUTION
FR,JB,SS,andMWdevelopedtheideasandmethodology;FR,JB,SS,andMWcollectedthedata;FRanalyzedthedata,withcontribu-tionsfromJB,LM,SS,andMW;FRledthewritingofthemanuscript
withsubstantialcontributionsfromJB,LM,SS,andMW.Allauthorsgavefinalapprovalforpublication.
DATA ACCE SSIBILIT Y
DatawillbemadeavailableintheDryadDigitalRepository.
ORCID
Felix Ratcliff http://orcid.org/0000-0002-3105-6264
R E FE R E N C E S
Aschmann,H.(1984).Arestrictivedefinitionofmediterraneanclimates-Bulletin de La Société Botanique de France. Actualités Botaniques,131(2-4), 21–30. https://doi.org/10.1080/01811789.1984.10826643
Baldwin,B.G.(Ed.)(2012).The Jepson manual: Vascular plants of California,(2nded.)Berkeley,CA:UniversityofCaliforniaPress.
Bartolome, J. W., Allen-Diaz, B. H., Barry, S., Ford, L. D., Hammond,M.,Hopkinson, P.,…White,M.D. (2014).Grazing for biodiversityinCalifornianMediterranean grasslands.Rangelands,36(5), 36–43.https://doi.org/10.2111/Rangelands-D-14-00024.1
Belsky,A. J.,Matzke,A.,&Uselman,S. (1999). Surveyof livestock in-fluencesonstreamandriparianecosystems in thewesternUnitedStates.Journal of Soil and Water Conservation,54,419–431.
Bestelmeyer,B.T.,Brown,J.R.,Havstad,K.M.,Alexander,R.,Chavez,G.,&Herrick,J.E.(2003).Developmentanduseofstate-and-transitionmodels for rangelands. Journal of Range Management, 56(2), 114.https://doi.org/10.2307/4003894
Bestelmeyer, B. T., Goolsby, D. P., & Archer, S. R. (2011). Spatial per-spectives in state-and-transition models: A missing link to landmanagement?Journal of Applied Ecology,48(3),746–757.https://doi.org/10.1111/j.1365-2664.2011.01982.x
Borcard, D., Gillet, F., & Legendre, P. (2011). Numerical ecology with R. New York, NY: Springer New York. Retrieved from http://link.springer.com/10.1007/978-1-4419-7976-6 https://doi.org/10.1007/978-1-4419-7976-6
Campbell, C. J., & Green, W. (1968). Perpetual succession ofstream-channel vegetation in a semiarid region. Journal of the Arizona Academy of Science, 5(2), 86–98. https://doi.org/10.2307/40024609
Caudle,D.,DiBenedetto, J.,Karl,M., Sanchez,H.,&Talbot,C. (2013).Interagency ecological site handbook for rangelands. United StatesDepartmentofAgriculture:NaturalResourcesConservationService.
DeCaceres,M.,&Legendre,P.(2009).Associations between species and groups of sites: Indices and statistical inference.RPackage.Retrievedfromhttp://sites.google.com/site/miqueldecaceres/.
Dibblee, T. W. (2005).Geologic Map of the Pastoria Creek Quadrangle. DibbleeGeologyCenterMap#DF-175.,SantaBarbaraMuseumofNaturalHistory.
Dibblee,T.W.(2008a).Geologic map of the Arvin and Tejon hills quadran-gles.DibbleeGeologyCenterMap#DF-416.,SantaBarbaraMuseumofNaturalHistory.
Dibblee, T.W. (2008b).Geologic map of the Neenach and willow springs quadrangles.DibbleeGeologyCenterMap#DF-383.,SantaBarbaraMuseumofNaturalHistory.
Dufrene,M.,&Legendre,P.(1997).Speciesassemblagesandindicatorspecies:Theneedforaflexibleasymmetricalapproach.Ecological Monographs,67(3),345–366.https://doi.org/10.2307/2963459
Gasith,A.,&Resh,V.H. (1999). Streams inmediterranean climate re-gions: Abiotic influences and biotic responses to predictable
12 | RATCLIFF eT AL.
seasonal events. Annual Review of Ecology and Systematics, 30(1),51–81.https://doi.org/10.1146/annurev.ecolsys.30.1.51
George,M.,Jackson,R.,Boyd,C.,&Tate,K.(2011).Conservation benefits of rangeland practices: Assessment, recommendations, and knowledge gaps. United StatesDepartment ofAgriculture,Natural ResourcesConservationService.
Jackson,R.D.,Bartolome,J.W.,&Allen-Diaz,B.(2002).Stateandtransi-tionmodels-responsetoESAsymposium:Thresholdsandnon-linearresponses inecosystems:Understanding, sustaining, and restoringcomplexrangelands.Bulletin of the Ecological Society of America,83,194–195.
James,G.,Daniela,W.,Trevor,H.,&Robert,T.(2013).An introduction to statis-tical learning Vol. 103. Springer texts in statistics.NewYork,NY:SpringerNew York. Retrieved from http://link.springer.com/10.1007/978-1-4614-7138-7.https://doi.org/10.1007/978-1-4614-7138-7
Kauffman, J. B., & Krueger,W. C. (1984). Livestock impacts on ripar-ian ecosystems and streamside management implications. A re-view. Journal of Range Management, 37(5), 430–438. https://doi.org/10.2307/3899631
Kondolf, G. M., Montgomery, D., Piegay, H., & Schmitt, L. (2003).GeomorphicClassificationofRiversandStreams.InG.M.Kondolf&P.Hervé(Eds.),Tools in fluvial geomorphology.Hoboken,NJ:JWiley.https://doi.org/10.1002/0470868333
Lichvar,R.W.,Banks,D.W.,Kirchner,W.N.,&Melvin,N.C.(2016).Thenationalwetlandplant list: 2016wetland ratings.Phytoneuron,30,1–17.
Lunt, I. D., Jansen, A., Binns, D. L., & Kenny, S. A. (2007). Long-Termeffects of exclusion of grazing stock on degraded herbaceousplantcommunities ina riparianeucalyptusCamaldulensis forest inSouth-EasternAustralia.Austral Ecology,32(8),937–949.https://doi.org/10.1111/j.1442-9993.2007.01782.x
Maechler, M., Rousseeuw, P., Struyf, A., Hubert, M., & Hornik, K.(2015).Cluster: Cluster analysis basics and extensions.Rpackagever-sion2.0.3.
McBride, J., & Strahan, J. (1984). Fluvial Processes and Woodland Succession Along Dry Creek, Sonoma County, California. InE.Richard(Ed.),Warner and Kathleen M.Hendrix:UniversityofCaliforniaPress.
McCune,B.,Grace,J.B.,&Urban,D.L.(2002).Analysis of ecological com-munities.GlenedenBeach,OR:MjMSoftwareDesign.
NationalCartographyandGeospatialCenter(NRCS)(2010).10 meter na-tional elevation data.FortWorth,TX:USGeologicalSurvey.
Oksanen, J., Blanchet,G., Kindt, R., Legendre, P.,Minchin, P.,O’Hara,R. B., … Wagner, H. (2015). Vegan: community ecology package. R package version 2.3-5. Retrieved fromhttp://CRAN.R-project.org/package=vegan.
Perevolotsky,A.,&Seligman,N.G.(1998).Roleofgrazinginmediterra-neanrangelandecosystems.BioScience,48(12),1007–1017.https://doi.org/10.2307/1313457https://doi.org/10.2307/1313457
RCore Team (2016).R: A language and environment for statistical com-puting. Vienna, Austria: R Foundation for Statistical Computing.Retrievedfromhttps://www.R-project.org/.
Ripley,B.(2016).Tree: Classification and regression trees.Rpackageversion1.0-37.Retrievedfromhttps://CRAN.R-project.org/package=tree.
Sabo,J.L.,Sponseller,R.,Dixon,M.,Gade,K.,Harms,T.,Heffernan,J.,…Welter, J. (2005). Riparian zones increase regional species rich-nessbyharboringdifferent,notmore,species.Ecology,86(1),56–62.https://doi.org/10.1890/04-0668
Sankey, J. B., Ralston, B. E., Grams, P. E., Schmidt, J. C., &Cagney, L.E. (2015). Riparian vegetation, Colorado River, and climate: Five
decadesofspatiotemporaldynamicsinthegrandcanyonwithriverregulation. Journal of Geophysical Research: Biogeosciences, 120(8),2015JG002991.https://doi.org/10.1002/2015jg002991
Spiegal,S.,Huntsinger,L.,Hopkinson,P.,&Bartolome,J.(2016).Rangeecosystems inecosystemsof california. InA.Harold (Ed.),Mooney and erika zavaleta.Oakland,CA:UniversityofCaliforniaPress.
Spiegal,S.,Larios,L.,Bartolome,J.,&Suding,K.N.. (2014).Grasslandsbiodiversityandconservationinachangingworld.InP.Mariotte&P.Kardol(Eds.),Environmental science, engineering and technology. New York,NY:NovaPublishers.
Stella, J. C., Battles, J. J.,McBride, J. R., &Orr, B. K. (2010). Riparianseedling mortality from simulated water table recession,and the design of sustainable flow regimes on regulated riv-ers. Restoration Ecology, 18(November), 284–294. https://doi.org/10.1111/j.1526-100X.2010.00651.x
Stringham,T.K.,Krueger,W.C.,&Shaver,P.L. (2003).Stateandtran-sition modeling: An ecological process approach. Journal of Range Management,56(2),106–113.https://doi.org/10.2307/4003893
Stringham,T.K.,Krueger,W.C.,&Thomas,D.R.(2001).Applicationofnon-equilibriumecologytorangelandriparianzones.Journal of Range Management,54(3),210–217.https://doi.org/10.2307/4003236
Stringham, T., & Repp, J. (2010). Ecological site descriptions:Considerationforripariansystems.Rangelands,32(6),43–48.https://doi.org/10.2111/Rangelands-D-10-00085.1
Tate,K.W.,Atwill,E.R.,Bartolome,J.W.,&Nader,G.(2006).SignificantEscherichia coli attenuation by vegetative buffers on annual grass-lands. Journal of Environmental Quality,35(3),795–805.https://doi.org/10.2134/jeq2005.0141
Vaghti,M.,&Greco,S.(2007).RiparianVegetationoftheGreatValley.InM.G.Barbour,T.Keeler-Wolf&A.A.Schoenherr(Eds.),Terrestrial vegetation of California [electronic Resource].Berkeley,CA:UniversityofCaliforniaPress,c2007.
Westoby,M.,Walker,B.,&Noy-Meir, I. (1989).Opportunisticmanage-mentforrangelandsnotatequilibrium.Journal of Range Management Archives,42(4),266–274.https://doi.org/10.2307/3899492
Winward, A. H. (2000).Monitoring the vegetation resources in riparian areas. Ogden, UT: US Department of Agriculture, Forest Service,Rocky Mountain Research Station. Retrieved from http://www.fs.fed.us/rm/pubs/rmrs_gtr047.pdf. https://doi.org/10.2737/RMRS-GTR-47
Zuur,A.F.,Ieno,E.N.,&Smith,G.M.(2007).Analysing ecological data. New York, NY; London: Springer. Retrieved from http://public.eblib.com/choice/publicfullrecord.aspx?p=371384. https://doi.org/10.1007/978-0-387-45972-1
SUPPORTING INFORMATION
AdditionalSupportingInformationmaybefoundonlineinthesup-portinginformationtabforthisarticle.
How to cite this article:RatcliffF,BartolomeJ,MacaulayL,SpiegalS,WhiteMD.Applyingecologicalsiteconceptsandstate-and-transitionmodelstoagrazedriparianrangeland.Ecol Evol. 2018;00:1–12. https://doi.org/10.1002/ece3.4057