Geoscience Reference
In-Depth Information
of.a.species.is.of.an.autocorrelative.nature.in.that.the.presence.of.a.species.in.a.spe-
ciic.site.in.one.way.or.another.depends.on.the.presence.of.that.species.somewhere.
else. nearby. in. space. and. time. (Brown. and. Lomolino. 1998;. Araújo. et. al.. 2005a)..
Thus,.in.order.to.have.truly.robust.validation.tests,.geostatistical.techniques.must.be.
developed.
If. validating. niche. models. in. the. present. time. seems. complex,. doing. so. for.
events. that. have. not. yet. occurred—as. do. the. models. that. use. climatic. conditions.
in. the. future—is. even. more. challenging.. Strictly. speaking,. distribution. models.
for. the. future. cannot. be. validated.. Probably. the. only. way. to. test. the. trustworthi-
ness. of. such. distribution. models. is. by. assessing. whether. the. modeling. exercise. is.
reliable.for.sets.of.data.from.the.recent.past.(decades).to.the.present..For.example,.
Araújo.et.al..(2005a).modeled.116.species.of.birds.in.the.United.Kingdom.for.two.
time. periods. (1968-1972. and. 1988-1991). in. which. both. species'. occurrences. and.
climatologies.were.chronologically.aligned..Then.they.generated.models.with.four.
methods.(Generalized.Linear.Models.[GLM],.Generalized.Additive.Models.[GAM],.
Classiication.Tree.Analysis.[CTA],.and.an.Artiicial.Neural.Network.[ANN]).using.
the. data. from. 1968-1972. and. projected. onto. climatic. conditions. for. 1988-1991.
and. validated. with. occurrences. associated. with. the. second. period.. Their. results.
showed. that. ANN. and. GAM. produced. projections. that. are. more. reliable. than. did.
the.other.methods..In.another.study,.Hijmans.and.Graham.(2006).produced.distri-
bution.predictions.for.the.past,.present,.and.future.with.four.niche.modeling.tech-
niques.(BIOCLIM,.DOMAIN,.GAM,.and.Maxent).and.compared.their.performance.
against.that.of.a.mechanistic.model..They.found.that.Maxent.produced.models.more.
similar.to.those.obtained.with.the.mechanistic.models.
In.summary,.model.projections.for.future.scenarios.are,.in.fact,.highly.variable.
and.may.or.may.not.be.reliable..Production.of.highly.reliable.and.statistically.signii-
cant.models.that.describe.present.relationships.is.not.enough.to.guarantee.that.mod-
els.for.the.future.will.also.be.reliable..Although.that.is.certainly.a.mandatory.irst.
step,.whenever.possible.it.is.greatly.valuable.to.cross-validate.models.with.recent-
past.data.to.establish.which.method.works.better.for.a.particular.case.
a
lTernaTiVe
a
pproaChes
for
p
rediCTing
s
peCies
'
d
isTribuTions
under
C
limaTe
C
hange
Ecological.Niche.Modeling.is.to.date.by.far.the.most.popular.approach.for.modeling.
species'. ranges. and. distributional. shifts. due. to. climate. change.. However,. it. is. not.
the. only. approach;. others. have. been. developed. under. a. more. knowledge-oriented.
paradigm..Mechanistic.or.process-based.models.(Figure.4.2).aim.to.(1).understand.
the.genetic,.physiological,.demographic,.or.behavioral.mechanisms.that.underlie.the.
way(s).in.which.environmental.factors.affect.the.performance.and.survival.of.indi-
viduals. and. populations,. and. (2). use. that. knowledge. to. infer. species'. geographic.
distributions.and.shifts.that.may.result.from.climatic.changes.(Crozier.and.Dwyer.
2006;.Helaouet.and.Beaugrand.2009).
Mechanistic.models.are.certainly.more.robust.than.ENMs,.but.have.the.dis-
advantage. that. they. require. greater. amounts. of. data.. In. addition. to. spatial. and.
