ArticlePDF Available

Abstract and Figures

Seedling survival is a limiting factor in arid-land restoration. We investigated how variation in the root traits of glasshouse-reared seedlings related to the field performance of different genotypes from two populations of Elymus elymoides (squirreltail), a common bunchgrass native to the Western United States. Seeds from 100 E. elymoides individuals were collected from two sites in northern Nevada. We planted offspring of these 100 individuals in the glasshouse to characterize 10-day root traits of each maternal family. Root traits of glasshouse-reared plants and seed size measures were correlated with the performance of siblings grown in field plots close to the seed collection sites. Seedling root traits were related to performance of siblings at both sites. We estimate that within-population variation in root traits was associated with a more than six- to nine-fold increase in seedling survival probability and a two-fold increase in height at the less productive site, and a two-fold increase in survival and 1.2-fold increase in size at the more productive site. At both sites, effects of root traits were complex, with extreme values of some traits favoured and intermediate values of other traits favoured. There is a recognized need to integrate understanding of plant functional traits into larger conceptual frameworks of ecological restoration. Here we show that within-population variation in a suite of root functional traits relates to large variation in seedling survival and size of an arid-land grass species, improving our understanding of how trait variation affects performance in the field. Understanding such variation may be used to positively impact restoration outcomes.
Content may be subject to copyright.
Please
cite
this
article
in
press
as:
Atwater,
D.
Z.,
et
al.
Seedling
root
traits
strongly
influence
field
survival
and
performance
of
a
common
bunchgrass.
Basic
and
Applied
Ecology
(2015),
http://dx.doi.org/10.1016/j.baae.2014.12.004
ARTICLE IN PRESS
BAAE-50848;
No.
of
Pages
13
Basic
and
Applied
Ecology
xxx
(2015)
xxx–xxx
Seedling
root
traits
strongly
influence
field
survival
and
performance
of
a
common
bunchgrass
Daniel
Z.
Atwatera,,
Jeremy
J.
Jamesb,
Elizabeth
A.
Legera
aDepartment
of
Natural
Resources
and
Environmental
Science,
University
of
Nevada,
Mail
Stop
186,
1000
Valley
Road,
Reno,
NV
89557,
USA
bSierra
Foothill
Research
and
Extension
Center,
University
of
California,
8279
Scott
Forbes
Road,
Browns
Valley,
CA
95918,
USA
Received
30
May
2014;
accepted
13
December
2014
Abstract
Seedling
survival
is
a
limiting
factor
in
arid-land
restoration.
We
investigated
how
variation
in
the
root
traits
of
glasshouse-
reared
seedlings
related
to
the
field
performance
of
different
genotypes
from
two
populations
of
Elymus
elymoides
(squirreltail),
a
common
bunchgrass
native
to
the
Western
United
States.
Seeds
from
100
E.
elymoides
individuals
were
collected
from
two
sites
in
northern
Nevada.
We
planted
offspring
of
these
100
individuals
in
the
glasshouse
to
characterize
10-day
root
traits
of
each
maternal
family.
Root
traits
of
glasshouse-reared
plants
and
seed
size
measures
were
correlated
with
the
performance
of
siblings
grown
in
field
plots
close
to
the
seed
collection
sites.
Seedling
root
traits
were
related
to
performance
of
siblings
at
both
sites.
We
estimate
that
within-population
variation
in
root
traits
was
associated
with
a
more
than
six-
to
nine-fold
increase
in
seedling
survival
probability
and
a
two-fold
increase
in
height
at
the
less
productive
site,
and
a
two-fold
increase
in
survival
and
1.2-fold
increase
in
size
at
the
more
productive
site.
At
both
sites,
effects
of
root
traits
were
complex,
with
extreme
values
of
some
traits
favoured
and
intermediate
values
of
other
traits
favoured.
There
is
a
recognized
need
to
integrate
understanding
of
plant
functional
traits
into
larger
conceptual
frameworks
of
ecological
restoration.
Here
we
show
that
within-population
variation
in
a
suite
of
root
functional
traits
relates
to
large
variation
in
seedling
survival
and
size
of
an
arid-land
grass
species,
improving
our
understanding
of
how
trait
variation
affects
performance
in
the
field.
Understanding
such
variation
may
be
used
to
positively
impact
restoration
outcomes.
Zusammenfassung
Das
Überleben
der
Sämlinge
ist
ein
limitierender
Faktor
bei
der
Renaturierung
in
trockenen
Gebieten.
Wir
untersuchten
wie
die
Variation
von
Wurzelmerkmalen
bei
im
Gewächshaus
gezogenen
Sämlingen
sich
zur
Freilandperformanz
von
unterschiedlichen
Genotypen
aus
zwei
Populationen
von
Elymus
elymoides,
einem
häufigen,
einheimischen
Tussockgras
des
Westens
der
Ver-
einigten
Staaten
verhält.
Samen
von
100
Individuen
von
E.
elymoides
wurden
auf
zwei
Flächen
in
Nord-Nevada
gesammelt.
Wir
pflanzten
die
Nachkommen
dieser
100
Individuen
im
Gewächshaus
aus,
um
nach
zehn
Tagen
die
Merkmale
der
Wurzeln
jeder
Mutterfamilie
zu
beschreiben.
Die
Wurzelmerkmale
der
Pflanzen
aus
dem
Gewächshaus
und
die
gemessenen
Samengrößen
wurden
mit
der
Performanz
von
verwandten
Individuen
korreliert,
die
auf
Freiflächen
nahe
den
Sammelstellen
wuchsen.
Die
Corresponding
author.
Present
address:
Department
of
Plant
Pathology,
Physiology,
and
Weed
Science,
Virginia
Tech,
435
Old
Glade
Rd.
0330,
Blacksburg,
VA
24061,
USA.
E-mail
address:
danatwater@gmail.com
(D.Z.
Atwater).
http://dx.doi.org/10.1016/j.baae.2014.12.004
1439-1791/©
2015
Published
by
Elsevier
GmbH
on
behalf
of
Gesellschaft
für
Ökologie.
Please
cite
this
article
in
press
as:
Atwater,
D.
Z.,
et
al.
Seedling
root
traits
strongly
influence
field
survival
and
performance
of
a
common
bunchgrass.
Basic
and
Applied
Ecology
(2015),
http://dx.doi.org/10.1016/j.baae.2014.12.004
ARTICLE IN PRESS
BAAE-50848;
No.
of
Pages
13
2
D.Z.
Atwater
et
al.
/
Basic
and
Applied
Ecology
xxx
(2015)
xxx–xxx
Wurzelmerkmale
der
Sämlinge
standen
auf
beiden
Flächen
in
Beziehung
zur
Performanz
der
Geschwister.
Wir
kalkulierten,
dass
die
Variation
der
Wurzelmerkmale
innerhalb
einer
Population
mit
einer
sechs-
bis
über
neunfachen
Zunahme
der
Überlebensrate
der
Sämlinge
verbunden
war
und
mit
einer
1.2-fachen
Größenzunahme
auf
der
produktiveren
Fläche.
Auf
beiden
Flächen
waren
die
Effekte
der
Wurzelmerkmale
komplex,
wobei
bei
manchen
Merkmalen
extreme
Werte,
bei
anderen
aber
intermediäre
Werte
begünstigt
wurden.
Es
besteht
ein
anerkannter
Bedarf
dafür,
das
Verständnis
von
funktionalen
Pflanzenmerkmalen
in
das
konzeptionelle
Bezugssystem
der
ökologischen
Renaturierung
zu
integrieren.
Hier
zeigen
wir,
dass
die
Variation
einer
Reihe
von
funktionellen
Wurzelmerkmalen
innerhalb
einer
Population
mit
der
großen
Variationsbreite
von
Überlebensrate
und
Größe
der
Sämlinge
einer
Grasart
der
Trockengebiete
zusammenhängt.
Dies
verbessert
unser
Verständnis
davon,
wie
diese
Variation
die
Performanz
im
Freiland
beeinflusst.
Und
das
Verständnis
derartiger
Variabilität
kann
genutzt
werden,
um
die
Ergebnisse
von
Renaturierungen
positiv
zu
beeinflussen.
©
2015
Published
by
Elsevier
GmbH
on
behalf
of
Gesellschaft
für
Ökologie.
Keywords:
Bromus
tectorum;
Elymus
elymoides;
Great
Basin;
Restoration;
Root
architecture;
Root
development;
Root
size;
Seed
mass
Introduction
Roots
are
a
major
area
of
interface
between
plants
and
their
environment.
Root
architecture,
i.e.
the
surface
char-
acteristics,
branch
morphology,
and
gross
topology
of
root
tissue
(after
Fitter,
1987),
is
known
to
influence
nutrient
uptake
(Busso,
Briske,
&
Olalde-Portugal,
2001),
stress
resis-
tance
(Rundel
&
Nobel,
1991),
interaction
with
competitors
(Casper
&
Jackson
1997),
and
interaction
with
soil
microor-
ganisms
(Hetrick,
1991;
Nibau,
Gibbs,
&
Coates,
2008;
also
reviewed
by
Hodge,
Berta,
Doussan,
Merchan,
&
Crespi,
2009).
While
simulation
models
have
shown
that
the
length,
depth,
and
arrangement
of
roots
can
affect
the
rate,
amount
and
location
of
nutrient
uptake
(e.g.
Somma,
Hopmans,
&
Clausnitzer,
1998),
experimental
studies
of
the
ecological
importance
of
root
form
are
difficult,
due
to
the
extreme
plas-
ticity
of
root
traits
and
to
the
practical
difficulty
of
directly
observing
processes
in
soil.
As
a
result,
we
understand
some
of
the
drivers
of
plastic
changes
in
root
form
and
we
can
associate
differences
in
root
architecture
with
changes
in
resource
capture
(see
reviews
by
Lynch,
1995;
Casper
&
Jackson,
1997;
Hodge
et
al.,
2009),
but
we
still
know
lit-
tle
about
how
variation
in
root
form
affects
plant
survival
and
function
(Lynch,
1995;
Pacheco-Villalobos
&
Hardtke,
2012).
Among
species,
variation
in
root
traits
is
believed
to
have
important
effects
on
the
ecology
of
plant
species.
This
is
certainly
the
case
for
plants
in
water-limited
environments,
where
adequate
root
proliferation
is
necessary
to
take
advan-
tage
of
scant
rain
during
the
growing
season
(Gregory
2008).
However,
roots
are
costly
to
produce,
and
overproduction
of
roots
could
increase
the
risk
of
water
loss
in
dry
soils.
Cor-
rectly
balancing
these
conflicting
pressures
is
thought
to
be
crucial
for
plants
growing
in
arid
environments
(Canadell
et
al.,
1996;
Jackson
et
al.,
1996;
Schulze
et
al.,
1996).
For
grasses
growing
in
arid
environment,
it
is
believed
that
pro-
liferation
of
shallow
roots
aids
water
capture
during
sporadic
rain
events,
although
the
exact
relationship
between
root
form
and
ecological
function
is
still
unknown
(Rundel
&
Nobel,
1991;
Donovan
&
Ehleringer,
1994;
Gibbens
&
Lenz,
2001).
Within-species,
most
studies
have
focused
on
the
impor-
tance
of
plastic
variation
in
root
form,
e.g.
in
response
to
varying
resource
abundance
(reviewed
by
Lynch,
1995;
Casper
&
Jackson,
1997).
However,
non-plastic,
herita-
ble
within-species
variation
in
root
form
is
also
known
to
exist,
particularly
for
crop
species
and
model
organisms
(reviewed
by
Pacheco-Villalobos
&
Hardtke,
2012).
In
crop
species,
root
traits
have
been
linked
to
performance
in
water-
limited
environments
(Manschadi,
Hammer,
Christopher,
&
deVoil,
2008).
In
model
organisms
like
Arabidopsis,
rice,
and
tomato,
rapid
progress
is
being
made
in
our
understanding
of
the
genetic
controls
over
root
architecture
in
these
species
(reviewed
in
Nibau
et
al.,
2008).
Much
less
is
known
about
how
within-species
variation
in
root
form
affects
the
per-
formance
of
wild
plants,
however
Rowe
and
Leger
(2011)
showed
that
invasion
by
Bromus
tectorum
was
associated
with
changes
in
a
large
number
of
root
traits
of
the
arid-
land
native
grass
Elymus
multisetus,
including
changes
in
root
length,
diameter,
and
branching
patterns.
This
variation
was
also
found
to
affect
competitive
performance
in
a
glasshouse
experiment.
Understanding
the
relationship
between
seedling
traits
and
field
performance
is
important
for
understanding
the
ecology
of
arid
plants,
but
it
may
also
provide
informa-
tion
that
is
important
for
land
stewardship.
For
example,
disturbance
by
fire,
climate
change,
and
invasion
represent
major
threats
to
Great
Basin
ecosystem
health
and
natural
resource
quality
(Reynolds
et
al.,
2007;
UNCCD,
2012).
Restoration
to
combat
natural
resource
loss
requires
the
use
of
appropriate
native
seed
sources,
but
how
best
to
identify
such
seed
sources
is
the
subject
of
a
large
and
long-
lasting
debate
in
restoration
ecology
(see
McKay,
Christian,
Harrison,
&
Rice,
2005;
Broadhurst
et
al.,
2008;
Vander
Please
cite
this
article
in
press
as:
Atwater,
D.
Z.,
et
al.
Seedling
root
traits
strongly
influence
field
survival
and
performance
of
a
common
bunchgrass.
Basic
and
Applied
Ecology
(2015),
http://dx.doi.org/10.1016/j.baae.2014.12.004
ARTICLE IN PRESS
BAAE-50848;
No.
of
Pages
13
D.Z.
Atwater
et
al.
/
Basic
and
Applied
Ecology
xxx
(2015)
xxx–xxx
3
Mijnsbrugge,
Bischoff,
&
Smith
2010).
It
is
widely,
but
not
universally,
believed
that
locally
accessed
seeds
are
supe-
rior
for
restoration
of
native
rangeland
habitat.
Unfortunately,
such
material
is
rarely
available
and
it
may
not
be
obtain-
able
following
severe
disturbance.
However,
it
may
also
be
possible
to
identify
surrogate
genotypes
for
restoration
by
matching
phenotypic
traits
to
the
environment
being
restored.
The
need
to
link
our
understanding
of
functional
trait
variation
to
larger
conceptual
frameworks
of
ecological
restoration
is
widely
acknowledged
(Funk,
Cleland,
Suding,
&
Zavaleta,
2008;
Cadotte,
Carscadden,
&
Mirotchnick,
2011;
Drenovsky
et
al.,
2012),
and
species
mixes
used
for
restoration
are
already
chosen
based
on
functional
traits
(Pywell
et
al.,
2003;
Kempel,
Chrobock,
Fischer,
Rohr,
&
van
Kleunen,
2013)
and
proximity
or
similarity
to
the
envi-
ronment
being
restored
(McKay
et
al.,
2005;
Noël
et
al.
2011).
It
may
also
be
possible
to
choose,
below
the
species
level,
optimal
genotypes
for
restoration
based
on
functional
traits.
Here,
we
investigate
the
role
that
variation
in
root
traits
plays
in
the
field
performance
of
Elymus
elymoides
(Raf.)
Swezey
(Poaceae;
commonly:
squirreltail),
an
economically
important
bunchgrass
that
is
common
in
sagebrush
steppe
rangeland
ecosystems
of
the
Great
Basin,
even
persisting
in
areas
that
are
densely
invaded
by
cheatgrass,
B.
tectorum
L.
We
assessed
root
traits
in
glasshouse-reared
E.
elymoides
seedlings
and
related
variation
in
root
traits
to
survival
and
growth
of
maternal
siblings
grown
in
a
natural
setting
at
two
sites
in
the
Northern
Great
Basin.
Because
seed
size
is
known
to
affect
seedling
growth,
especially
at
the
early
stages
of
plant
establishment
(Baker,
1972;
Buckley,
1982;
Jurado
&
Westoby,
1992),
we
also
accounted
for
variation
in
seed
mass
and
how
it
may
have
affected
root
trait
expression
and
field
performance.
While
many
studies
have
investigated
root
dif-
ferences
among
species,
fewer
have
examined
root
variation
at
the
population
and
individual
level,
and
we
know
of
no
such
study
linking
individual
differences
in
root
architec-
ture
with
field
performance
of
a
wild
plant.
Here,
we
focus
on
native
grass
seedling
performance
in
B.
tectorum-invaded
sagebrush-steppe
sites,
as
understanding
the
dynamics
that
promote
native
bunchgrass
establishment
in
these
commu-
nities
is
of
extreme
management
concern
(e.g.
Davies
et
al.,
2011).
Root
traits
such
as
root
length,
diameter,
and
branching
have
been
shown
to
play
a
role
in
the
ability
of
Elymus
species
to
compete
with
invasive
annuals
(Rowe
&
Leger
2011),
and
they
are
also
believed
to
impact
survival
in
the
harsh
environment
of
the
Northern
Great
Basin.
We
hypothesized
that
root
length,
root
diameter,
and
degree
of
root
forking
of
glasshouse-reared
plants
would
be
important
predictors
of
survival
and
growth
of
field-grown
seedlings
after
one
grow-
ing
season,
and
further
hypothesized
that
similar
traits
would
be
beneficial
in
two
similar
study
sites.
We
tested
this
hypoth-
esis
by
measuring
root
traits
of
10-day
E.
elymoides
seedlings
grown
in
the
greenhouse
and
related
root
trait
variation
to
ger-
mination,
survival,
and
growth
of
siblings
grown
in
the
field.
We
focused
on
first-year
growth,
as
this
is
believed
to
be
the
most
critical
phase
in
the
life
cycle
of
perennial
arid-land
grasses
(James,
Svejcar,
&
Rinella,
2011).
Materials
and
methods
Study
species
and
research
design
E.
elymoides
is
a
cool-season
perennial
bunchgrass
that
is
common
throughout
the
Great
Basin
and
Mountain
West
of
the
United
States.
It
is
an
important
component
of
cold
desert
systems
in
the
Great
Basin
(McInnis
&
Vavra,
1987)
due
to
its
ability
to
withstand
annual
grass
invasion
(Booth,
Caldwell,
&
Stark,
2003)
and
persist
after
fire
(Wright
&
Klemmedson,
1965;
Young
&
Miller,
1985).
Elymus
species
reproduce
pri-
marily
via
self-fertilization
(Smith,
1944),
resulting
in
high
homozygosity
and
high
genetic
similarity
among
maternal
siblings
(Knapp
&
Rice,
1996;
Wilson,
Kitzmiller,
Rolle,
&
Hipkins,
2001).
In
July
of
2012,
E.
elymoides
seeds
were
collected
sep-
arately
from
fifty
plants
(hereafter,
“families”)
at
each
of
two
sites
(100
seed
families
total):
Paradise
Valley
(hereafter,
“Paradise”,
41.3425N,
117.6012W)
and
Orovada
(41.5419N,
117.7824W).
These
sites
were
located
in
Northern
Nevada,
U.S.A.,
selected
because
they
are
similar
sites
separated
by
roughly
27
km
and
because
they
are
representative
of
the
burned
and
subsequently
Bromus-invaded
sagebrush
steppe
communities
that
are
common
targets
of
restoration
efforts
in
the
Great
Basin.
Paradise
is
a
slightly
cooler
and
drier
site
on
average
(climate
averages
from
1981
to
2009,
Paradise:
annual
precipitation,
247
mm,
mean
warmest
month
tem-
perature:
22.8 C;
Orovada:
annual
precipitation,
254
mm,
mean
warmest
month
temperature:
23.3 C;
PRISM
climate
group,
http://prism.oregonstate.edu).
During
the
course
of
our
study,
which
occurred
during
a
series
of
drought
years
in
Nevada,
precipitation
was
low
but
similar
between
the
two
sites
(Paradise:
growing
season
precipitation
(Dec.
1
to
June
15.)
=
77.9
mm;
Orovada:
74.0
mm).
However,
Orovada
had
fewer
but
larger
rain
events
and
was
cloudier
(Par-
adise:
mean
solar
irradiance
=
220.8
mol
m2s1;
Orovada:
176.9
mol
m2s1).
As
a
result
the
soil
stayed
wet
much
later
into
the
season
at
Orovada
(see
Appendix
A:
Fig.
1).
Seeds
from
some
families
were
limited,
and
were
care-
fully
divided
between
a
screening
phase,
where
we
grew
plants
in
glasshouse
to
quantify
seedling
root
traits,
and
field
uses,
where
siblings
of
these
seeds
were
planted
in
the
wild.
To
avoid
an
experimental
bias
towards
highly
fecund
geno-
types,
we
designed
our
experiment
in
such
a
way
that
even
the
least
fecund
family
produced
enough
seed
to
use
in
our
experiment
(about
45
seeds).
Five
seeds
from
each
of
the
100
families
were
planted
into
66
mL
RLC4
Cone-tainers
(Stuewe
&
Sons)
filled
with
a
50/50
mix
of
local
top
soil
and
30-grit
sand
and
grown
for
10
days
post-emergence
in
the
glasshouse
for
root
trait
analysis
(n
=
5).
This
time
frame
and
seed
count
was
selected
because
previous
stud-
ies
(Rowe
&
Leger,
2011)
indicated
that
root
traits
at
10
Please
cite
this
article
in
press
as:
Atwater,
D.
Z.,
et
al.
Seedling
root
traits
strongly
influence
field
survival
and
performance
of
a
common
bunchgrass.
Basic
and
Applied
Ecology
(2015),
http://dx.doi.org/10.1016/j.baae.2014.12.004
ARTICLE IN PRESS
BAAE-50848;
No.
of
Pages
13
4
D.Z.
Atwater
et
al.
/
Basic
and
Applied
Ecology
xxx
(2015)
xxx–xxx
days
had
a
stronger
correlation
with
plant
performance
in
a
glasshouse
competition
study
than
root
traits
measured
at
later
time
points.
Additionally,
desiccation
is
known
to
be
a
major
cause
of
seedling
mortality
(Fenner,
1987),
and
early
development
of
root
tissue
is
believed
to
be
crucial
in
pro-
moting
survival
at
this
important
life
history
stage
(James
et
al.,
2011).
Above
and
below
ground
tissues
were
harvested,
with
root
systems
gently
washed
from
planting
media.
Intact
roots
were
scanned
and
measured
using
WinRHIZO
image
analysis
software
(Regular
Edition;
Regent
Instruments
Inc.,
2013).
After
scanning,
roots
and
leaves
were
separated,
dried,
and
weighed.
Roots
were
scored
for
total
length,
surface
area,
volume,
mean
diameter,
specific
root
length
(SRL:
root
length/volume),
number
of
root
tips
(hereafter,
“tips”),
root
mass
ratio
(RMR:
root
mass/plant
total
mass).
These
traits
were
selected
because
they
were
found
to
be
important
for
E.
elymoides
performance
in
glasshouse
studies
(Rowe
&
Leger,
2011).
We
also
measured
days-to-emergence
for
each
seedling,
a
trait
important
for
field
performance
of
E.
ely-
moides
seedlings
(Kulpa
&
Leger,
2013).
A
random
sample
of
approximately
50
seeds
was
weighed
for
each
family.
Forty
seeds
from
each
family
were
mounted
on
toothpicks
with
Tightbond
II
glue
(to
ease
identification
after
emergence
in
the
field)
and
planted
into
ten
blocks
at
the
home
site
for
that
family.
We
compared
several
glues
in
a
glasshouse
environment,
and
found
that
this
glue
did
not
affect
germina-
tion
timing,
growth
rates,
or
size
of
grass
seedlings
relative
to
unglued
controls
(Wehan
et
al.
unpublished
data).
Seeds
were
planted
on
October
30
and
November
1,
2012,
prior
to
the
arrival
of
germinating
rains.
At
each
site,
blocks
were
located
inside
a
100
m2subplot
within
a
1
ha
experimen-
tal
exclosure.
The
exclosures
had
been
sprayed
the
previous
spring
with
herbicide
to
reduce,
but
not
eliminate,
seed
pro-
duction
of
resident
annuals,
and
the
plant
communities
at
the
time
of
the
experiment
consisted
of
scattered
native
perennial
grasses
(primarily
Poa
secunda)
and
abundant
cover
of
the
exotic
invader
B.
tectorum.
Seeds
were
planted
directly
into
the
existing
plant
communities
to
simulate
restoration
into
natural
conditions.
Because
not
enough
seed
was
available
to
plant
seeds
from
each
family
at
both
sites,
seeds
were
planted
only
into
their
site
of
origin.
Two
replicates
from
each
family
were
planted
per
block,
for
a
total
of
20
replicates
per
family
per
site.
In
total,
1000
seeds
were
planted
at
each
site.
Emer-
gence
was
monitored
after
germinating
rains
and
every
three
to
four
weeks
throughout
winter,
ending
in
June
when
plants
began
to
senesce,
at
which
point
the
height
of
each
surviving
plant
was
recorded.
To
quantify
productivity
in
each
site,
the
aboveground
biomass
of
the
co-occurring
vegetation
in
each
plot
was
collected
in
June,
and
dried
and
weighed.
Statistical
analysis
We
used
principal
components
analysis
(PCA)
to
investi-
gate
the
correlations
among
root
traits
and
guide
the
selection
of
a
set
of
candidate
variables
for
further
analysis
(Jolliffe,
2002).
This
analysis
was
done
in
R
using
the
“psych”
package
(Revelle,
2013)
with
data
from
both
sites
combined.
Our
goal
was
to
identify
a
subset
of
variables
with
correlation
coefficients
<0.70,
in
order
to
avoid
collinearity
problems
in
subsequent
models.
We
chose
a
single
variable
from
each
component
axis
that
loaded
strongly
onto
that
axis.
There
were
strong
correlations
among
root
length,
surface
area
and
volume,
the
primary
contributors
to
PC1
(|r|
>
0.88;
see
Appendix
A:
Fig.
2),
and
between
root
average
diameter
and
SRL
(r
=
0.92),
the
primary
contributors
to
PC2.
These
traits
were
not
strongly
correlated
with
other
measurements
(number
of
tips,
RMR,
or
days-to-emergence;
|r|
<
0.40),
except
for
a
correlation
between
number
of
root
tips
and
root
length
(Paradise:
r
=
0.669;
Orovada:
r
=
0.749;
see
Appendix
A:
Figs.
2
and
3)
and
between
root
length
and
RMR
at
Par-
adise
(r
=
0.601).
Thus,
for
the
final
analysis
we
retained
root
length,
SRL,
number
of
tips,
RMR,
and
days-to-emergence,
noting
that
some
of
these
traits
were
strongly
correlated
with
traits
that
were
not
retained
for
further
analysis.
That
is,
the
effects
of
root
length
can
also
be
interpreted
as
potential
effects
of
surface
area
or
volume,
and
effects
of
SRL
can
also
be
interpreted
as
potential
effects
of
root
diameter
on
plant
performance.
We
also
included
seed
mass
as
a
covariate
and
tested
its
correlations
with
the
selected
root
traits.
Heritabilities
of
all
traits
except
seed
mass,
which
was
measured
on
a
per-family
rather
than
per-seed
basis,
were
estimated
using
generalized
linear
mixed
models
with
site
as
a
fixed
factor,
where
random
variance
associated
with
family
was
divided
by
the
total
variance
to
produce
a
broad-
sense
heritability
estimate
under
glasshouse
conditions.
We
note
that
our
method
of
estimating
broad-sense
heritability
assumes
very
high
genetic
similarity
among
maternal
sib-
lings.
This
is
likely
a
safe
assumption,
as
Elymus
species
have
high
self-fertilization
rates
(Smith,
1944).
We
estimated
sig-
nificance
for
random
effects
of
family
using
the
R
package
“RLRsim”
(Scheipl,
Greven,
&
Kuechenhoff,
2008).
Both
sites
were
analysed
separately.
We
analysed
the
effects
of
root
traits
on
plant
performance
using
multi-model
inference.
We
calculated
corrected
Aikake
information
criterion
(AICc)
scores
for
generalized
linear
mixed
models
(GLMMs;
using
the
R
package
“lmer”;
Bates,
Maechler,
&
Bolker,
2013)
with
seed
mass,
root
length,
SRL,
number
of
tips,
RMR,
and
days-to-emergence
as
continuous
predictors,
and
block
and
family
as
random
factors.
Predic-
tors
were
standardized
to
facilitate
interpretation
of
quadratic
terms
and
model
comparison.
Models
were
constructed
with
every
possible
combination
of
predictors,
except
that
second-
order
effects
were
not
allowed
unless
both
main
effects
were
also
present,
and
quadratic
terms
were
not
allowed
unless
linear
terms
were
also
present.
The
response
variable
was
one
of
three
metrics
of
plant
performance:
emergence
prob-
ability,
survival
probability
through
the
first
growing
season,
and
height
after
the
first
growing
season.
We
then
performed
model
averaging,
in
which
we
averaged
the
parameter
estimates
of
the
top
subset
of
models
(those
with
AICc
4),
weighted
according
to
their
AICc
scores.
This
AICc
Please
cite
this
article
in
press
as:
Atwater,
D.
Z.,
et
al.
Seedling
root
traits
strongly
influence
field
survival
and
performance
of
a
common
bunchgrass.
Basic
and
Applied
Ecology
(2015),
http://dx.doi.org/10.1016/j.baae.2014.12.004
ARTICLE IN PRESS
BAAE-50848;
No.
of
Pages
13
D.Z.
Atwater
et
al.
/
Basic
and
Applied
Ecology
xxx
(2015)
xxx–xxx
5
threshold
was
selected
because
it
represents
a
middle
ground
among
commonly
suggested
thresholds
(usually
between
2
and
10;
Grueber,
Nakagawa,
Laws,
&
Jamieson,
2011).
This
technique
is
useful
when
trying
to
understand
the
effects
of
a
large
number
of
partially
correlated
predictor
variables,
as
it
is
less
biased
and
more
accurate
than
performing
stepwise
regressions
or
choosing
to
analyse
only
the
top
model
using
information
criteria
(Burnham
&
Anderson,
2002).
When
performing
model
averaging,
researchers
must
con-
tend
with
the
fact
that
a
different
subset
of
predictors
is
present
in
each
of
the
models
being
compared.
Thus,
one
must
decide
how
to
average
across
predictors
that
are
absent
from
some
of
the
models
being
considered.
There
are
two
approaches
to
solve
this
problem.
One
approach,
“zero
aver-
aging,”
assigns
a
parameter
estimate
of
zero
in
every
model
in
which
a
given
predictor
is
missing,
and
it
includes
those
zeroes
in
the
model
averaging.
For
example
if
root
length
is
included
in
three
out
of
the
five
top
models
of
survival
proba-
bility,
the
parameter
estimates
in
the
three
models
containing
root
length
would
be
averaged
with
zeroes
in
the
two
models
lacking
root
length,
to
calculate
the
zero-averaged
estimate.
Another
approach,
“natural
averaging”,
simply
ignores
mod-
els
that
lack
the
predictor
being
averaged,
instead
averaging
across
only
those
models
where
the
given
predictor
appears.
In
the
example
given
above,
the
naturally-averaged
estimate
would
simply
be
the
weighted
average
of
the
parameter
estimates
in
the
three
models
containing
root
length.
Natural-
averaged
parameters
are
thought
to
be
better
for
assessing
the
importance
of
single
parameters,
while
zero-averaged
param-
eters
are
thought
to
be
more
useful
for
comparing
the
relative
importance
of
multiple
parameters
(Burnham
&
Anderson,
2002;
Grueber
et
al.,
2011).
Model
averaging
also
produces
an
estimate
of
parameter
importance,
which
refers
to
the
pro-
portion
of
the
top
subset
of
models
that
contain
the
parameter
of
interest,
weighted
according
to
the
AICc
values
of
the
models
containing
that
parameter.
A
greater
value
means
that
the
parameter
is
present
in
more
models
and/or
models
that
are
more
informative
(i.e.
that
have
a
lower
AICc
value).
Confidence
intervals
can
be
calculated
for
naturally-averaged
parameters,
but
at
present
methods
for
this
do
not
exist
for
zero-averaged
parameters.
Model
averaging
was
done
using
the
“MuMIN”
package
for
R
(Barton,
2013).
Model
averaging
can
easily
handle
complex
effects,
such
as
interactions
between
terms
and
quadratic
predictors,
how-
ever
it
also
tends
to
produce
mathematically
complex
models
because
it
retains
every
term
that
occurs
at
least
once
in
the
list
of
top
models.
For
example,
if
root
length
had
a
linear
effect
on
survival
in
19
of
20
top
models
and
a
quadratic
effect
on
survival
in
one
of
20
top
models,
the
averaged
model
would
describe
a
weak
quadratic
effect
of
root
length.
This
would
not
affect
our
ultimate
conclusions
but
it
would
make
interpreta-
tion
of
parameter
estimates
more
laborious
and
less
intuitive.
To
avoid
this
complication
and
to
facilitate
downstream
analyses,
we
screened
models
for
non-significant
quadratic
effects
before
entering
quadratic
effects
into
the
initial
model
pool
(α
=
0.100;
P-values
estimated
using
the
Satterthwaite
approximation;
R
package
“lmerTest”;
Kuznetsova
et
al.
2013).
In
the
models
described
above,
seed
mass
was
allowed
to
have
direct
effects
on
field
performance.
However,
because
seed
mass
can
affect
expression
of
root
traits,
it
may
also
have
had
indirect
effects
on
performance
via
its
effect
on
root
traits.
For
example,
large
seeds
could
directly
improve
survival
by
providing
more
resources
to
a
seedling
but
also
by
enabling
greater
root
growth.
To
estimate
the
importance
of
indirect
influences
of
seed
mass,
we
used
a
path
analytic
approach.
The
indirect
effect
of
seed
mass
via
each
root
trait
was
estimated
by
multiplying
the
correlation
between
seed
mass
and
a
given
root
trait
with
the
β-coefficient
describing
that
root
trait’s
linear
effect
on
performance
(from
the
models
described
above).
We
note
that
this
analysis
is
not
possible
on
quadratic
effects,
because
they
change
in
value
and
sign
with
different
root
trait
values.
Therefore,
we
did
not
perform
this
analysis
on
predictors
with
quadratic
terms,
and
we
did
not
attempt
to
calculate
total
indirect
effects
of
seed
mass.
We
also
note
that
the
correlations
compared
in
this
method
were
obtained
in
different
environments
(the
glasshouse
and
the
field)
and
that
this
method
may
only
produce
a
rough
approximation
of
the
actual
indirect
effects
of
seed
mass.
We
performed
these
and
all
other
analyses
in
R
Version
2.15.1
(R
Core
Team,
2012).
Graphs
were
produced
using
the
“ggplot2”
package
(Wickham,
2009).
Results
Differences
between
sites
Seed
mass
varied
greatly
between
sites,
with
Orovada
seeds,
on
average,
being
over
twice
the
size
of
Paradise
seeds
(Fig.
1;
t103 =
29.01,
P
<
0.001).
Root
traits
also
dif-
fered
between
sites,
with
seedlings
from
Orovada
having
longer,
thicker
roots
(lower
SRL)
with
more
tips
(Fig.
1;
|t|104 >
4.371,
P
<
0.001).
Orovada
seedlings
also
invested
relatively
less
biomass
into
root
growth
(lower
RMR;
t104 =
3.333,
P
=
0.001)
with
shorter
time
to
emergence
(t104 =
8.097,
P
<
0.001).
Productivity
of
co-occurring
plants
also
varied
between
the
two
sites,
with
Orovada
hav-
ing
about
three
times
greater
productivity
during
the
study
than
Paradise
(Paradise:
26.2
g
m2during
the
study;
Orovada:
79.6
g
m2;
t18 =
7.639,
P
<
0.001).
We
attribute
this
difference
in
productivity
to
the
greater
soil
moisture
at
Orovada.
Inheritance
of
root
and
seed
traits
Root
traits
measured
in
the
glasshouse
differed
sig-
nificantly
among
families,
with
broad-sense
heritabilities
between
0.110
and
0.214
(P
0.038;
see
Appendix
A:
Table
1)
for
all
traits
other
than
number
of
root
tips,
for
which
ran-
dom
effects
of
family
were
too
weak
to
be
estimated
by
our
mixed
model
analysis,
as
effects
were
small
relative
to
overall
variance.
Please
cite
this
article
in
press
as:
Atwater,
D.
Z.,
et
al.
Seedling
root
traits
strongly
influence
field
survival
and
performance
of
a
common
bunchgrass.
Basic
and
Applied
Ecology
(2015),
http://dx.doi.org/10.1016/j.baae.2014.12.004
ARTICLE IN PRESS
BAAE-50848;
No.
of
Pages
13
6
D.Z.
Atwater
et
al.
/
Basic
and
Applied
Ecology
xxx
(2015)
xxx–xxx
Fig.
1.
Family
averages
for
root
traits
in
this
study.
Each
point
is
a
family
mean
for
the
two
wild-collected
sites.
Analyses
presented
indicate
comparisons
between
sites.
Table
1.
Effects
of
seed
and
root
traits
on
field
performance,
determined
using
generalized
linear
mixed
models.
Standardized
parameters
estimated
from
the
naturally-averaged
model
(βNA)
and
zero-averaged
model
(βzero)
are
shown.
Values
range
from
-1
(maximum
negative
effect)
to
1
(maximum
positive
effect).
Because
quadratic
coefficients
can
be
difficult
to
interpret
we
direct
readers
to
Fig.
2
to
see
these
effects
plotted,
but
we
note
that
negative
quadratic
effects
suggest
intermediate
values
are
favoured.
Estimates
of
statistical
significance
for
naturally
averaged
effects
are
indicated
with
asterisks
(P
0.100; *P
0.050; ** P
0.010).
The
relative
parameter
importance
(imp.)
in
the
zero-averaged
models
is
also
shown.
This
statistic
gives
the
proportion
of
models
containing
that
parameter,
weighted
by
the
AICc
values
of
the
models
in
which
the
parameter
appears.
A
higher
value
means
that
the
parameter
is
in
more
models
and/or
models
with
a
lower
AICc
score,
to
a
maximum
of
one.
A
blank
row
means
that
parameter
was
not
present
in
any
of
the
models
with
AICc
4
or
because
it
was
a
pre-emptively
omitted
quadratic
term
(see
text).
DTE
refers
to
days-to-emergence.
Emergence
Survival
Height
Paradise
Orovada
Paradise
Orovada
Paradise
Orovada
βNA βzero Imp.
βNA +
Imp.
βNA βzero Imp.
βNA βzero Imp.
βNA βzero Imp.
βNA βzero Imp.
Seed
Mass
0.111
0.051
0.46
0.1440.094
0.65
0.087
0.030
0.35
0.1220.089
0.73
0.145*0.050
0.34
0.0880.011
0.13
Length
0.344*0.301
0.87
0.101
0.033
0.33
0.339*0.270
0.80
0.107
0.067
0.63
0.188*0.099
0.53
Length20.164*0.103
0.63
SRL
0.227*0.207
0.91
0.094
0.038
0.41
0.324*0.317
0.98
0.065
0.013
0.20
0.0880.011
0.13
Tips
0.247*0.203
0.82
0.161
0.089
0.56
0.181
0.099
0.55
0.108
0.074
0.69
0.126*0.006
0.05
0.0960.016
0.17
Tips20.032
0.006
0.18
0.096
0.029
0.31
0.124** 0.006
0.05
RMR
0.168
0.086
0.51
0.005
0.001
0.22
0.2310.171
0.74
0.024
0.001
0.03
0.1480.021
0.14
RMR20.018
0.002
0.13
0.011
0.001
0.13
DTE
0.064
0.020
0.31
0.091
0.037
0.41
0.028
0.005
0.17
0.104
0.050
0.48
0.1310.015
0.11
Please
cite
this
article
in
press
as:
Atwater,
D.
Z.,
et
al.
Seedling
root
traits
strongly
influence
field
survival
and
performance
of
a
common
bunchgrass.
Basic
and
Applied
Ecology
(2015),
http://dx.doi.org/10.1016/j.baae.2014.12.004
ARTICLE IN PRESS
BAAE-50848;
No.
of
Pages
13
D.Z.
Atwater
et
al.
/
Basic
and
Applied
Ecology
xxx
(2015)
xxx–xxx
7
Influence
of
seed
and
root
traits
on
emergence,
survival,
and
<