Patternsofspeciesrichnessand

endemism

SergeiVolis

TashkentBotanicalGarden

Whypatternsofspeciesrichnessand

endemismareimportant?

?narrowendemicsareusuallyrareandthereforepotentiallythreatened

?speciesrichnessandendemismarestronglyrelated

?However,centresofhighspeciesrichnessandcentresofhighendemicityarenotalways

concordantwhichsuggeststhatbothspeciesrichnessandendemismmustbeconsideredinconservationdecision.

Myersetal.Nature2000

Currentlyrecognizedglobalhotspotsofplantendemism,whicharedefinedashaving>1500endemicplantspecies

Harrison&Noss2017

Huangetal.Biol.Cons.2016

Huangetal.Front.PlSc.2022

Noroozietal.Biol.Cons.2019

ConservationpriorityzonesinMadagascar

Kremenetal.Science2008

?Weightedendemism(WE)isproportionaltotheinverseofaspecies’rangeandsumoftheweightsindicatesthedegreeofendemismamongthespeciesoccurringinacell

?Specieswithmorerestrictedrangesareassignedhighweight,andthosewithlargerrangesareassigneda

progressivelylowerweights

?Correctedweightedendemism(CWE)dividesWEbythespeciesrichnessofacell

?CWEcorrectsforadependenceonspeciesrichness

becausecellswithhighnumberofspeciesareexpectedtohavelargenumberofendemicsbychancealone,thusnotrepresentingtruecentersofendemism

WE=1/6+1/5=0.36

WE=1/1+1/1=2

WE=1/1+1/1+1/1+1/1+1/1+1/1=6

?Weightedendemism(WE)isproportionaltotheinverseofaspecies’rangeandsumoftheweightsindicatesthedegreeofendemismamongthespeciesoccurringinacell

?Specieswithmorerestrictedrangesareassignedhighweight,andthosewithlargerrangesareassigneda

progressivelylowerweights

?Correctedweightedendemism(CWE)dividesWEbythespeciesrichnessofacell

?CWEcorrectsforadependenceonspeciesrichness

becausecellswithhighnumberofspeciesareexpectedtohavelargenumberofendemicsbychancealone,thusnotrepresentingtruecentersofendemism

CWE=(1/1+1/1+1/1+1/1+1/1+1/1)/6=1

CWE=(1/56+1/56+…)/6=0.018

CWE=(1/1+1/1)/2=1

?CWEcanbeinterpretedastheaverageperspeciesrangerestrictiontoacell

?CWEcanbeinterpretedasaproportionofspeciesrangesrestrictedtoacell

?CWE=0.2meansthatonaverage20%ofthespeciesinacellarerestrictedtothatcell,or,alternatively,thata

cellrepresents20%ofthespeciesrange

?CWEisanalogoustobetadiversitydefinedasadegreeofdifferentiationinspeciescompositionbetween

locations/habitats

Laffan&Crisp2003

Thequalityofbiodiversityassessmentsusingtheabovemetricsdependsonthequalityofthedatasetsandthe

methodusedtodeterminethespeciesdistributions

point-to-gridmaps

?assemblespeciesoccurrencerecordswithingridcellsandcountthenumberofspeciesobservedineachcell.Thismethodhasthe

advantageofnotextrapolatingdata,butitusuallysuffersfroma

largenumberoffalseabsences(asituationwhenaspeciesmayincorrectlyappeartobeabsentfromaparticularlocation,whereinfactitdoesoccur,simplybecausethatlocationhasneverbeen

surveyed).

?usefulforcoarse-scaledconservationassessments,wherefalse-absencesarerarelyaproblembecausethespatialunitsintheseassessmentarelargeenoughtocontainsurveydata

?however,lessusefulforfine-scaledregionalplanningbecausewhenspatialunitsaresmall,usuallyonlyasmallproportionofthemare

surveyed.

?almostinevitableinsufficientsamplingeffortcausestheaccuracyofthismethodtodecreasewithanincreaseincellresolution.

Effectofsamplingeffort

Oliveiraetal.Diver.Distr.2016

speciesdistributionmodelling(SDM)

?allowsnotonlyfillinggeographicalgapsininformationonspeciesdistributions,butalsoreducingthefrequencyoffalse-absences

inherentinspeciesoccurrencedatasets

?althoughhasaknowntendencytooverestimatetruespeciesrangesiswidelyacceptedasausefultoolinconservationnotonlyfor

individualspeciesassesments,butalsoforelucidatingspatialpatternsofbiodiversity

DistributionofAntechinusagilis

Amboni&LaffanInt.JGeogr.Inf.Sci.2012

DispersalabilitiesOverestimation

Speciesinteractions

SDM:

9bioclimatic

1topographic

6vegetationvariablesBinarypredictions

Maximizingthesumofspecificity-sensitivity

thresholdofprobability

r

Amboni&LaffanInt.JGeogr.Inf.Sci.2012

?thepredictedspecieshabitatsuitabilitymapsgeneratedbySDMareoverlaid(stacked)toproducediversitymaps.

?suchaggregatingindividual-speciesvaluespercellforspecies

inhabitingthesamegeographicalregion,iscalledstackedspeciesdistributionmodelling(S-SDM)

BuildinganS-SDM

?selectionofanSDMmethod

generaladditivemodels(GAM),

generalizedlinearmodels(GLM),

multivariateadaptiveregressionsplines(MARS),

classificationtreeanalysis(CTA),

generalizedboostedmodels(GBM),

maximumentropy(MAXENT),

artificialneuralnetworks(ANN),

randomforests(RF),

supportvectormachines(SVM)

?conversionofeachsingleSDMfromacontinuoussuitabilitymaptoabinarypresence/absencemapusingathresholdcriterion(or

alternatively,thesumofthecontinuousvalues)

?thechoiceofoneoranotherSDMmethodandthethresholdagainsttheuseofcontinuousvaluessignificantlyaffecttheS-SDMoutcome

Point-to-gridmapsvsS-SDM

?Thechoiceofthetwogeneralapproacheswillinfluencethespeciesrangemapsandresultingbiodiversitymetrics,andthusthe

recommendedconservationpriorities

?AlthoughS-SDMhasbeeninuseformorethanadecade,its

efficiencyasatoolindescribingpatternsofspeciesrichnesshasbeentestedinalimitednumberofstudies

?Graham&Hijmans2006werethefirsttocomparethepoint-to-gridspeciesrichnessmapswiththoseproducedfromSDMpredictions

andshowedthatSDM-basedmapsover-predictedspeciesrichness,butthat”likelythatthepoint-to-gridmethoddoesnotcaptureactualspeciesrichnessunlessaratherlargegridcellsizeisused”.

SpeciesrichnessofamphibianandreptilespeciesinCalifornia

Pearson’sr

Point-to-gridDistributionmodel

0.393

25km

0.622

50km

Graham&HijmansGl.Ecol.Biog.2006

RichnessmapsformarsupialsinAustralia

BiodiverseMaxent

Pearson’sr

0.40

0.45

0.45

Hotspots-top5%cells

Amboni&LaffanInt.JGeogr.Inf.Sci.2012

SpeciesrichnessofoaksinMexico

Hernández-Quirozetal.Biod.&Cons.2018

?SDMoverpredictsspeciesrichnessatallresolutions,whilepoint-to-gridmapsunderpredictit

?Unfortunately,analogouscomparisonsofendemismmapshavenotbeenreportedintheliterature.

?Tofillthisgap,Iconductedastudycomparingthesetwotypesofmapsforthreemonocotgenera(Allium,TulipaandEremurus).

?Separateanalysesconductedforgroupsofcloselyrelatedtaxaareadvantageousoverusageofphylogeneticallyunrelatedspecies

becausetheformerareexpectedtorespondsimilarlyto

environmentalfactors,thusreducingthecomplexityoftheresponse.

Studygoals

?comparisonofmapsofspeciesrichness,WEandCWEusingpoint-to-gridvs.S-SDMmethods(thelatterproducedusingeitherbinaryorcontinuousvaluesof

predictedhabitatsuitability)

?EffectofsamplingdensityonS-SDM-producedmetricsofbiodiversity

?Previousstudiesthatmanipulatedsamplesizesweresingle-speciesSDMs

?Inthepresentstudy,samplesizeswere

manipulatedtoimitatepoorsamplingdensitybythinningspeciesoccurrencesusingrandom

numbersimulatortonomorethan10occurrencesperspecies

?patternsofspeciesrichnessandendemism

producedbyS-SDMfromfullvsreduceddatasetwerecompared

ssdmprogram(Schmittetal.2017)

MAXENTalgorithm

pSSDMandbSSDMstackingalgorithms

S-SDMmapsofspeciesrichness

producedfromoccurrencedatafor63Uzbekendemicplantspeciesby9

SDMalgorithmswithdefaultsettingsandpSSDMstackingalgorithm.

No.ofspecies

60

50

40

30

20

10

0

Allium

05101520304050>50

9

8

7

6

5

4

3

2

1

0

Eremurus

05101520304050>50

Occurrencerecords

8Tulipa

7

6

5

4

3

2

1

0

05101520304050>50

Allium

Eremurus

Tulipa

Results

?ThepatternsofspeciesrichnessandWEin

points-to-gridands-sdmmapswereverysimilarforallthreegenera

?But