Content uploaded by Dave Barnard
Author content
All content in this area was uploaded by Dave Barnard on Jun 20, 2019
Content may be subject to copyright.
LETTER Transient population dynamics impede restoration and may
promote ecosystem transformation after disturbance
Robert K. Shriver,*
1
Caitlin M.
Andrews,
1
Robert S. Arkle,
2
David M. Barnard,
2
Michael C.
Duniway,
3
Matthew J.
Germino,
2
David S.
Pilliod,
2
David A. Pyke,
4
Justin L. Welty,
2
and John B.
Bradford,
1
Abstract
The apparent failure of ecosystems to recover from increasingly widespread disturbance is a global
concern. Despite growing focus on factors inhibiting resilience and restoration, we still know very
little about how demographic and population processes influence recovery. Using inverse and for-
ward demographic modelling of 531 post-fire sagebrush populations across the western US, we
show that demographic processes during recovery from seeds do not initially lead to population
growth but rather to years of population decline, low density, and risk of extirpation after distur-
bance and restoration, even at sites with potential to support long-term, stable populations.
Changes in population structure, and resulting transient population dynamics, lead to a >50%
decline in population growth rate after disturbance and significant reductions in population den-
sity. Our results indicate that demographic processes influence the recovery of ecosystems from
disturbance and that demographic analyses can be used by resource managers to anticipate eco-
logical transformation risk.
Keywords
big sagebrush, demography, drylands, ecosystem transformation, regime shifts, resilience, restora-
tion, transient population dynamics.
Ecology Letters (2019)
INTRODUCTION
Climate change, altered disturbance regimes, and other anthro-
pogenic impacts are transforming ecosystems across the globe,
and the rate of ecosystem change is expected to increase over
the next century (Scheffer et al. 2001; Nolan et al. 2018; Rata-
jczak et al. 2018). Understanding the drivers of disturbance
and the mechanisms that inhibit or enable resilience following
disturbance have been a central focus of basic and applied
ecology during the past 40 years (Holling 1973). Ecologists
have identified a number of important abiotic (e.g. climate)
and biotic (e.g. invasive species) changes and feedbacks that
prevent the recovery of foundational plant species after distur-
bance and drive regime shifts, even after ecological restoration
is attempted (Suding et al. 2004; Suding & Hobbs 2009). Less
attention, however, has focused on understanding how changes
in populations following disturbance affect recovery trajecto-
ries. It is often implicitly assumed that in the absence of
changes in climate, invasive species, or other abiotic/biotic fac-
tors, disturbed populations should naturally recover (i.e.
increase in cover and density), given propagules are available.
However, the role of demographic and population processes in
determining recovery and resilience of foundational plant spe-
cies are rarely quantified (but see James et al. 2011; Larios
et al., 2017; Montero-Serra et al. 2018; Caughlin et al. 2019).
Disturbed plant populations are considerably altered in
their size structure from a pre-disturbance population. This is
especially true for long-lived species. Most plants exhibit
changes in vital rates (individual survival, fecundity, and
growth) as they develop and grow (Jones et al. 2014). Conse-
quently, perturbations in size structure can lead to transient
population dynamics, where the near-term dynamics of a pop-
ulation differs substantially from the long-term growth rate
that would be expected from a population if it were left undis-
turbed and allowed to reach its stable size distribution (SSD;
defined by the dominant Eigenvalue and corresponding right
Eigenvector of a matrix population model) (Caswell 2001;
Mcdonald et al. 2016). In other words, the rate of population
growth or decline is a function of population vital rates and
size structure. In the context of ecological resilience and
restoration, populations that begin as seeds after disturbance
may not recover but rather go through several years or dec-
ades of population declines as the population undergoes a
demographic transition from high mortality, low fecundity
seedlings and juveniles to larger plants that have lower mor-
tality and higher fecundity. This can lead to long periods of
low population density where species are functionally absent
and at-risk of extirpation from environmental or demographic
stochasticity, even if the population has the potential for long-
term growth once established (Fig. 1, Iles et al. 2016; Albrecht
et al. 2019). By providing a feedback mechanisms that func-
tionally removes foundational species from ecosystems after
disturbance, transient population dynamics alone could lead
to ecosystem regime changes that persist at management-
1
U.S. Geological Survey,Southwest Biological Science Center,2255 N Gemini
Rd, Flagstaff, AZ, USA
2
U.S. Geological Survey,Forest and Rangeland Ecosystem Science Center,970 S
Lusk St, Boise, ID, USA
3
U.S. Geological Survey,Southwest Biological Science Center,2290 Resource
Blvd, Moab, UT, USA
4
U.S. Geological Survey, Forest and Rangeland Ecosystem Science Center, 3200
SW Jefferson Way, Corvallis, OR, USA
*Correspondence: E-mail: rshriver@usgs.gov
Published 2019. This article is a U.S. Government work and is in the public domain in the USA
Ecology Letters, (2019) doi: 10.1111/ele.13291
relevant timescales (years, decades) or indefinitely (Hastings
2016; Hastings et al. 2018).
The widespread loss of sagebrush habitats is one of the lar-
gest restoration challenges in the United States. Historically,
habitats dominated by big sagebrush (Artemisia tridentata
Nutt.; hereafter sagebrush) covered ~620 000 km
2
(c. 8%) of
the contiguous United States (Davies et al. 2011). But, a legacy
of human activity, spread of invasive species, and an increase
in wildfire frequency and extent have resulted in 50–60% of
the historic sagebrush range being dominated by invasive
annual plants (West 2000). The inability of sagebrush to
resprout after fire (or develop a fire-resistant seed bank) have
driven practitioners to implement some of the most extensive
restoration actions in terrestrial ecosystems: seeding of up to
100 000s hectares annually (Pilliod et al. 2017; Copeland et al.
2018). Concerns about the decline of sagebrush obligate species
have made sagebrush restoration a top priority for many pub-
lic land management agencies, resulting in significant invest-
ment. For example, the 2017 U.S. Bureau of Land
Management (BLM) budget requested $79.2 million for the
restoration and conservation of sage-grouse and sagebrush
habitats alone (BLM 2016). But attempts to re-establish sage-
brush at disturbed sites have had mixed results at best. Some
sites rebound after the one-time reseedings, but many more
sites are remarkably resistant to the re-establishment of sage-
brush populations (Knutson et al. 2014). Where restoration
treatments are unsuccessful, treated areas often shift to grass-
lands, sometimes persistently (Chambers et al. 2014). Our
ability to explain mechanisms driving disparate outcomes has
been hampered by a lack of knowledge about the basic ecology
of sagebrush. One of the key pieces of missing information is
an understanding of the life-history and demography of sage-
brush, and how this influences its recovery dynamics across
broad environmental gradients. Remarkably, despite the large
investments made in restoration, to our knowledge no struc-
tured population model has ever been developed for sagebrush
(although see Tredennick et al. 2016; Kleinhesselink & Adler
2018 for population models based on cover, and Dalgleish
et al. 2011 for a structured model of a big sagebrush congener).
Here, we investigate whether transient population dynamics
substantially alter population growth rates after disturbance,
impede resilience and restoration, and in turn lead to ecosys-
tem transformation. We use an innovative approach to quan-
tify the vital rates and population dynamics of sagebrush at
531 sites across the Great Basin where seeding occurred after
wildfire. We overcome intense data-requirements traditionally
associated with demographic modeling (measuring individual
vital rates over several years and sites) by inferring vital rates
from changes in population size structure measured at field
sites, 2–36 years after seeding. This inverse modelling
approach allows us to quantify how vital rates and population
dynamics vary across large landscapes in a way that would
not be possible with traditional demographic methods. With
this approach, we quantify resilience based on recovery times
of population growth rate, density, and plant cover following
disturbance. We hypothesise that changes in size structure
associated with disturbance leads to transient population
declines and reduced density and cover after seeding, even in
populations with potential to grow in the long-term.
METHODS
Data
From 2014–2016, we collected population height distribution
data once at 531 Great Basin sites that had burned and were
subsequently reseeded by the BLM (see Fig. S13 for map and
Fig. S14 for distribution of years from seeding to measure-
ment). Of the 531 sampled sites, 370 sites were selected at ran-
dom. The remaining 161 sites were identified by participating
BLM field offices as areas where sagebrush establishment
likely occurred. Specific sampling locations within these
burned areas were chosen randomly. We included BLM
selected areas because low establishment of sagebrush at a site
is somewhat common, and these non-random sites provided
additional data for model parameterisation. Inclusion of non-
random sites did not change the overall conclusion of the
strength and impact of transient effects (Figs S19, S20).
At each site, we sampled size structure data along three
50 m transects of varying width, meant to maximise the accu-
racy and efficiency of measuring the density of each size class.
The width of transects began at a maximum of 6 m, but were
sequentially reduced to 4, 2, and 1 m if observers initially
expected to capture more than 20, 50, 70 individual plants in
the 6, 4, 2 m widths respectively. Sagebrush individuals were
binned in six height classes: 0–5 cm, 5–15 cm, 15–30 cm, 30–
75 cm, 75-120 cm, 120 +cm. Height distributions were
Density
Small
Large
Size Class
(a) (b) (c) (d)
Figure 1 How demographic processes can lead to transient population
dynamics and ecosystem transformation. After disturbance and seeding,
populations are dominated by a large number of smaller individuals that
germinate (a), likely far from their pre-disturbance or expected,
asymptotic size structure. Because survival and reproduction often
increase as a function of an individual’s size, disturbance could lead to
periods of transient population declines as the population undergoes a
demographic transition from low survival and fecundity small individuals
(b and c), to higher survival, higher fecundity large individuals (d).
Population declines driven by transient changes in population size
structure could lead to observed regime shifts: low population density or
extirpation, even in populations that may have been stable or growing
before disturbance or would be expected to grow in the long-term
(asymptotic dynamics). Code for figure model available in supplement
(Fig. 1 R Code) for readers to explore the role of vital rates in driving
transient dynamics
Published 2019. This article is a U.S. Government work and is in the public domain in the USA
2Shriver et al. Letter
converted to canopy cover assuming that sagebrush plants are
perfectly spherical (height =diameter). The largest size class
was assumed to end at 165 cm based on a subset of data in
which height to the nearest centimeter was measured.
Although less coarse height classes may have provided more
refined demographic inference, these data were originally
intended to quantify sagebrush habitat status, with the idea of
using them for demographic inference only coming after mea-
surement. We compiled seeding rate data on populations from
the Land Treatment Digital Library (LTDL; Pilliod & Welty
2013). The LTDL contains the best available records on BLM
restoration treatment in the Great Basin (Pilliod et al. 2017).
The bulk seeding rate (by weight) applied at each site was
extracted from LTDL records. When data were available, the
bulk weight was converted to pure live seed (i.e. viable seed
weight; PLS) using a site-specific value for the seed applied.
When a site-specific PLS rate was not available, the median
percent PLS for each subspecies across all sites was applied
(12.3% wyomingensis, 16% vaseyana, 13.6% tridentata).
Finally, the total weight of PLS was converted to number of
seeds using the average seed weight (2.5 mg per seed) across
subspecies-cytotypes (Richardson et al. 2015).
Demographic modeling
One of the major challenges in applying demographic models
to landscape-scale questions is the intense data requirements.
Traditionally, parameterising demographic models requires
sampling of individual-level vital rates over several years. This
is prohibitive for modelling demographic processes over vast
landscapes. We overcome this challenge by applying an under-
utilised approach for fitting structured population models
using time-series of population size structure. This approach
uses population-level data (i.e. the number and size of individ-
uals) to infer underlying vital rates and their drivers (Doak &
Morris 1999), and has been successfully applied to infer one
or more vital rates in diverse species (Hooten et al. 2007; Loso
& Doak 2006; Ghosh et al. 2012; Shriver et al. 2012). Because
inference happens at the population-level, Ghosh et al. (2012)
argued that this approach is likely to lead to more accurate
inference and predictions of population dynamics than tradi-
tional approaches of measuring and scaling-up individual-level
vital rates (Crone et al. 2013).
We developed a Bayesian modelling framework by mod-
elling the counts of individuals as a negative binomial,
ci;dNegative Binomial ni;d
tad;j
ð1Þ
where ci;dis the count of individuals in size class i(i=1...6)
in site d(d=1...531).n
i;d
tadis the mean occurrence rate,
where ni;d
tis the density (i.e. area standardised) by site and
size class in the year in which data were collected,
t, which
varies by site, and adis a scaling term to adjust the density to
the search area for each site. Finally, jis a parameter describ-
ing the dispersion of the data.ni;d
tis calculated based on a
matrix population model where
nd;t¼Ad;tnd;t1ð2Þ
here we use matrix notation to indicate that nd;tis a 6 91 vec-
tor containing the density of individuals in all size classes in
year t(t=1...
t), and Ad;tis a 6 96 state-transition matrix
describing how individuals move from one state to another in
a given year. Ad;tis defined in terms of its individual matrix
elements ai;jwhich are a function of underlying vital rates:
ai;j¼gi;jpj
zfflffl}|fflffl{
½3:1
þfjgi;1pi
zfflfflfflfflffl}|fflfflfflfflffl{
½3:2
ð3Þ
where the first term [3.1] describes the state-transitions of
existing plants: gi;jis the transition rate (i.e. growth or die-
back) from size jto i;pjis the survival probability of individu-
als in j. The second term [3.2] describes recruitment of new
individuals. fjare the seeds produced last year by existing
plants which germinate in the current year. Plants can grow
rapidly upon germination in the spring, before the mid-late
summer dry period. Accordingly, all new plants germinate
into the smallest size class and then grow (gi;1Þbefore their
survival probability is determined. First-year survival (piÞis
then based on the size class reached after initial growth.
Because sagebrush is widely recognised to have little to no
interannual seed survival (0-10% annual seed survival), a
seedbank is not included in the model (Wijayratne & Pyke
2012; Schlaepfer et al. 2014). Note, each element ai;j,and its
associated vital rates, are also indexed by site, d, and time, t,
but we have removed these indices from eq. 3 for clarity.
We use a parametric approach to fit vital rates as a function
of plant size, a year-specific climate effect, and site-specific
random effects (Loso & Doak 2006). For growth transitions,
the rate of transitioning to any size class i, given that the indi-
vidual is already in size class j,gi;j;d;t,is a normalised Gaussian
kernel.
gi;j;d;t¼ki;jðÞ¼Nðmijlj;d;t;r2Þ=X
6
i¼1
Nðmijlj;d;t;r2Þð4Þ
lj;d;t¼b0;dþb1;dmjþb2Moistured;tð5Þ
b0;dNormal b0;s2
0
ð6Þ
b1;dNormal b1;s2
1
ð7Þ
where bs are regression parameters, miis the midpoint size
of an individual in size class i, and b0;dx007E;and b1;dare
site-specific random effects for the intercept and size effect.
Moisture availability is known to be a critical climate vari-
able controlling sagebrush dynamics. Based on previous anal-
ysis (Shriver et al. 2018), we include average annual soil
moisture (VWC) in the spring period (DOY 70-100) to cap-
ture relevant interannual climate variability. Random site
effects summarise other average climate and site effects (e.g.
edaphic, sub-species) not captured by soil moisture. Using a
partial-pooling approach maximised the use of our data
across sites, while still permitting sites to vary in estimated
vital rates and dynamics. These differences among sites in
parameters and covariates that reflect abiotic and biotic fac-
tors allows us to overcome assumptions of traditional
chronosequence studies (Johnson & Miyanishi 2008).
Published 2019. This article is a U.S. Government work and is in the public domain in the USA
Letter Transient dynamics impede restoration 3
Likewise, survival is as follows:
logitðpj;d;tÞ¼b3;dþb4;dmjþb5Moistured;tð8Þ
b3;dNormalðb3;s2
3Þð9Þ
b4;dNormal b4;s2
4
ð10Þ
And, annual reproduction (i.e. seed production from previ-
ous year that then goes on to germinate in the current year) is
only a function of plant size and site random effects and does
not vary by year.
logðfj;dÞ¼b6;dþb7;dmjð11Þ
b6;dNormal b6;s2
6
ð12Þ
b7;dNormal b7;s2
7
ð13Þ
In practice, this parametric approach resembles the process
of fitting a continuous-state integral projection model (IPM)
(Easterling et al. 2000), but given our large, discrete size
classes the interpretation of model parameters is not always
analogous to an IPM. By using parametric equations and par-
tially pooling data across sites this approach allows us to infer
population dynamics and vital rates across all years and sites
given our available data. Finally, all bparameters were given
non-informative priors, Normal(0,5
2
) (see Table S1 for priors).
The model was initialised in the first growing season after seed-
ing with fjequal to the seeding rate in each site (see Data). The
likelihood of population size structure in the year of measure-
ment (eq. 1) was then evaluated from simulations in each Hamil-
tonian Monte Carlo (HMC) step using the proposed parameters.
Computation
Models were fit in a Bayesian framework using HMC in Stan.
Three parameter chains were run for 2000 iterations with a
1000 iteration warm-up period. HMC algorithms require far
fewer iterations to explore posterior parameter space than tradi-
tional Markov Chain Monte Carlo methods. Note that
although we present random effects using a centred parameteri-
sation for easier interpretation, the model was fit using a mathe-
matically equivalent non-centred parameterisation that is more
computationally efficient in Stan (Monnahan et al., 2017).
Parameter convergence was monitored visually and with con-
vergence statistics (R-hat). Information on model fit and valida-
tion using simulation, predictive checks, and sensitivity analyses
can be found in the supplemental information (Figs. S1-S12).
Demographic calculations
We used two metrics to summarise population results. First,
the stochastic population growth rate (i.e. geometric mean)
for each population from the first growing season following
seeding to the year of field measurement ksðÞ
.
ksðÞ¼exp Pt¼1ln kd;t
tmax
ð14Þ
Under asymptotic conditions kd;tis the dominant eigenvalue
of the matrix population model for site din year t.In
observed conditions, kd;tis Nd;t/Nd;t1, where Nd;tis the pre-
dicted total population density of population in site din year
tbeginning at the field seeding rate. Posterior predictions of
ksðÞ are made using full uncertainty from underlying parame-
ters. Posterior mean 95% CI are presented. We also simulated
annual observed population growth rate for 38 years after
seeding (i.e. Fig. 3c,d). We used the soil moisture conditions
at each site from seeding to field measurement and then ran-
domly drew an annual soil moisture condition from the same
site-specific empirical distribution each year to complete the
38-year time series.
Similarly, we quantified the strength of transient dynamics
by comparing observed population density (N
obs
) assuming
populations began as first year recruits from observed seeding
rates and asymptotic population density (N
asymp
) assuming
the same population density began at SSD (Iles et al. 2016).
The density of a population at the time of field observation
(N
obs
) was simulated using the seeding rate, posterior parame-
ter estimates, and soil moisture data. Posterior mean popula-
tion density in the year of field measurement was used as
N
obs
. The purpose of N
asymp
is to provide a reference value
indicating expected population density without transient
dynamics (i.e. at SSD). At SSD, population growth is equiva-
lent to the dominant eigenvalue of a transition matrix describ-
ing the population, which in our case is stochastic and varies
year-to-year. To capture the impact of stochasticity but elimi-
nate any effects of transient dynamics we calculated the
asymptotic population growth over the period from seeding to
field observations at each site (d)as Q
t¼2
kd;t, where kd;tis the
dominant eigenvalue of the transition matrix for site din year
t. We then multiplied by the mean posterior population den-
sity for each site in year 1 (i.e. N
obs
in year 1, the number of
seeds that survive the first growing season). We use the num-
ber of surviving individuals after the first year, rather than the
seeding rate because individual seeds are never explicitly
included in population density totals. Using the seeding rate
in the first year as the starting population density would
inflate the total population decline and impact of transient
dynamics.
Syntactical Note: Following the common notation of the
population dynamics literature, we use the term ‘observed’ to
distinguish predictions of population sizes and growth rates in
non-asymptotic conditions (i.e. beginning at field seeding rate)
from ‘asymptotic’ dynamics at SSD, but both ‘observed’ and
‘asymptotic’ conditions are derived from model predictions.
We use the term ‘field’ to indicate seeding rates and size struc-
tures derived from actual measured field data.
RESULTS
Posterior predictive checks indicate our model provides a
good fit to field measured size structure data (Figs S3, S4). In
addition, to ensure the validity of our demographic inferences,
we fit the model to simulated data from known parameter val-
ues. The model performed well with simulated data and recov-
ered the ‘true’ parameters values, with reasonably high
accuracy in most cases (Figs S1, S2).
Inferred annual survival probability and recruitment rate
increased dramatically with an individual’s size (Fig. 2). On
Published 2019. This article is a U.S. Government work and is in the public domain in the USA
4Shriver et al. Letter
average across all sites and years, annual survival probability
of plants 5–15 cm in height was 8%. Individuals in the largest
size classes (75–120 and 120–165 cm) had survival probabili-
ties approaching 100% (Fig. 2a). Similarly, larger plants were
predicted to produce far more seed and thus produce more
recruits. Plants in the 75–120 cm size class were, on average,
expected to produce ~0.73 recruits per year (Fig. 2b). In con-
trast, it would take an average of nearly 150 individuals in the
5–15 cm size class to produce one new recruit. Both growth
rates and survival probabilities increase with soil moisture
(Table S1). Estimates of asymptotic and transient kwere most
sensitive to changes in parameters relating plant size to vital
rates (b1,b4,b7; Fig. S11), while the rate at which small indi-
viduals grow to larger size classes was the vital rate with the
greatest impact on k(Figs S5–S10). Responses to soil mois-
ture and site-specific random effects led vital rates to vary
substantially across sites and years on average and as a func-
tion of plant size (Fig. S15). Higher soil moisture also con-
tributed to higher survival probabilities, growth rates and
recruitment rates at higher elevation (Figs S16, S17). Posterior
parameter estimates and credible intervals are available in
Table S1.
Population density declined in the first years after seeding,
starting at an average of ~40 established individuals per 50
m
2
, and declining to ~5 individuals per 50 m
2
after 10 years
(Fig. 3a). Low population sizes (~5 individuals per 50 m
2
) per-
sisted for nearly 15 years before population recovery began.
After 20 years average population density had not yet reached
the average density measured by Knutson et al. (2014), 32.7
per 50 m
2
, in undisturbed sites across the Great Basin.
Despite high initial population density, total sagebrush cover
did not begin to increase substantially on average until
15 years after disturbance (Fig. 3b). Population growth rates
immediately following disturbance were >50% lower than
long-term rates (Fig. 3c). Population growth rates increased
rapidly following disturbance, but many populations were pre-
dicted to have declined (k<1) for years to decades after seed-
ing as population size structure matured (Fig. 3c).
By ~20 years after seeding, most populations with long-term
potential to grow had reached k≥1 (Fig. 3d).
To determine the impact of changes in size structure on
population growth rates, we compared mean stochastic popu-
lation growth rate (ksðÞ
) assuming: (1) populations were at
SSD (asymptotic conditions); and (2) populations were begin-
ning at the observed seeding rate (observed conditions).
Results indicate consistent and substantial differences in ksðÞ
between asymptotic and observed populations. Specifically,
40% of populations had asymptotic ksðÞ ≥1 (based on poste-
rior mean ksðÞ
), while 94% of populations had their upper
95% CI for asymptotic ksðÞ ≥1 from the year of seeding to
measurement (Fig. 4a, S18a). However, in populations begin-
ning from seed the number of growing populations was cut
by 88%. Only 5% of sites had observed ksðÞ ≥1, and only
17% of populations had observed ksðÞ with their upper 95%
CI ≥1 from the year of seeding to measurement (Fig. 4b,
S18b). Because environmental gradients tend to vary consis-
tently across elevation in the Great Basin (e.g. average soil
moisture increases with elevation, Fig. S17), elevation is often
used as an important proxy for restoration planning and
implementation. Thus, we present estimates of ksðÞ across an
elevation gradient. Both asymptotic (P<0.0001) and
observed population (P<0.0001) growth rates declined on
average at lower elevations.
Finally, we quantified the importance of transient popula-
tion dynamics in driving observed population density.
Observed population densities (N
obs
, recovering from seed)
were on average >90% lower than asymptotic population
density (N
asymp,
the same population density beginning at
0.00
0.25
0.50
0.75
1.00
0−5 5−15 15−30 30−75 75−120 120−165
Height class (cm)
Survival Probability
1e−03
1e−01
1e+01
1e+03
0−5 5−15 15−30 30−75 75−120 120−165
Height class (cm)
Recruitment Rate
0.00
0.25
0.50
0.75
1.00
0−5 5−15 15−30 30−75 75−120 120−165
Height class (cm)
Growth Rate
(a) (b) (c)
Figure 2 Average annual vital rates across 531 big sagebrush populations. Values shown in box and whisker plots are the time-averaged vital rates for (a)
Survival, (b) Recruitment, (c) Growth across all years from the first growing season following seeding to the year of field measurement for each population
using each year’s posterior mean vital rate estimate. (b) Recruitment rates are the number of seeds produced by individuals of each size class that leadto
new plants that survive their first growing season and can vary from 0 to ∞. (c) Growth rates are the rate at which individuals in that size class transition
to any larger size class. Middle line in box indicates median, box spans the 25–75% inner quartile range (IQR), and upper and lower whisker reach
1.5*IQR
Published 2019. This article is a U.S. Government work and is in the public domain in the USA
Letter Transient dynamics impede restoration 5
0−5
5−15
15−30
30−75
75−120
120−165
0246
Density (plants 50m−2)
Height (cm)
Year 1
0−5
5−15
15−30
30−75
75−120
120−165
0246
Density (plants 50m−2)
Year 5
0−5
5−15
15−30
30−75
75−120
120−165
0246
Density (plants 50m−2)
Ye ar 1 0
0−5
5−15
15−30
30−75
75−120
120−165
0246
Density (plants 50m−2)
Ye a r 2 0
(a)
0
20
40
60
0102030
Population Density (plants 50m−2)
(b)
0
10
20
30
40
50
0102030
% Cover
(c)
0.1
0.3
1.0
0102030
Years since seeding
Observed λ
(
d
)
0.00
0.25
0.50
0.75
1.00
0 102030
Years since seeding
Proportion λ>1
Figure 3 Time series of inferred big sagebrush size structure and population dynamics after seeding at 531 sites, demonstrating population declines from
transient dynamics and slow recovery after disturbance. Barplots show the mean number of individuals in each size class across all populations 1, 5, 10 and
20 years after seeding. (a) mean posterior population density from the first growing season after seeding to the time of observation for each population
(blue) and the average across all populations (black). Red dashed line represents average density of sagebrush (32.7 per 50 m
2
) in unburned sites across the
Great Basin from Knutson et al. (2014). (b) mean posterior canopy cover from the first growing season after seeding to the time of observation for each
population (blue) and the average across all populations (black). In both (a) & (b) average line ends at 23 years because <5% of total populations extend
beyond this point. (c) Simulation results showing mean posterior population growth rates (k) for each population (blue) after seeding. Rates above red line
(k= 1) lead to growing populations, and below are declining populations. (d) Proportion of populations that have k1 over 38 years after seeding. Mean
(black line) and 95% CI (grey) are shown. Y-axis in (a) is truncated to only cover 0–60 for magnification.
Published 2019. This article is a U.S. Government work and is in the public domain in the USA
6Shriver et al. Letter
SSD) (Fig. 5). Similar to ksðÞ
, we investigated how observed
and asymptotic population density varied across an eleva-
tional gradient and found the reductions in population density
were greatest at low elevations (P<0.0001).
DISCUSSION
Our results indicate that changes in population size structure
caused by disturbance, and resulting transient population
dynamics, greatly impede the resilience of sagebrush and help
explain observed patterns of restoration failure and ecosystem
transformation. As hypothesised, we estimate that transient
dynamics led to reduced population growth rate and an aver-
age >90% decline in population density; this decline was
greatest at low elevations where post-fire sagebrush restora-
tion has historically been most difficult (Davies et al. 2011).
Likewise, our analyses suggest that populations typically con-
tinue to decline for years to decades after disturbance and
Growing populations
Declining populations
(a)
0.2
0.5
1.0
1.5
800 1200 1600 2000 2400
Elevation
Asymptotic λ
(b)
0.2
0.5
1.0
1.5
800 1200 1600 2000 2400
Elevation
Observed λ
Figure 4 Comparison of big sagebrush stochastic population growth rates in 531 populations across the Great Basin under (a) long-term, asymptotic
conditions, and (b) observed, post-disturbance conditions , illustrating lower growth rates as a result of transient population dynamics. Changes in size
structure from SSD to the observed seeding rate result in the number of growing populations being cut from 212 to 25, an 88% reduction. Points
correspond to posterior mean values and error bars cover the 95% CI. Green indicate populations predicted to grow (posterior mean ksðÞ ≥1) in
asymptotic and observed conditions, blue are those predicted to decline (ksðÞ <1) in asymptotic and observed conditions, and red are predicted to grow in
asymptotic conditions, but decline with changes in size structure in observed conditions. A corresponding figure using overlap of the upper 95% CI with
ksðÞ ≥1 (rather than posterior mean) is available in supplemental material (Fig. S18)
0.1
1.0
5.0
10.0
20.0
50.0
100.0
200.0
800 1200 1600 2000 2400
Elevation (m)
Relative Population Size (%, Nobs/Nasymp)
Figure 5 Negative impact of transient population dynamics on density of 531 big sagebrush populations depicted across an elevation gradient. Predicted
population density at the time of field observation (plants per 50m
2
) assuming population began as first year recruits from observed seeding rate (N
obs
)
relative to the same starting population density under asymptotic conditions (N
asymp
) across an elevation gradient
Published 2019. This article is a U.S. Government work and is in the public domain in the USA
Letter Transient dynamics impede restoration 7
seeding, increasing the risk that populations become locally
extinct and greatly slowing the recovery of density and cover
for populations that do persist. Importantly, these estimated
reductions in population density and growth following distur-
bance were driven by feedbacks from vital rates and popula-
tion structure. Although other factors (e.g. invasive species)
influence vital rates and thus the likelihood of recovery, our
results indicate that nearly all sites that may be capable of
supporting mature populations in the long-term (based on
their underlying vital rates) actually decline once population
size structure is changed by disturbance (Fig. 4). Thus, tran-
sient dynamics are likely to help drive low abundance and
cover after disturbance and seeding, leading sagebrush to be
functionally (or actually) extirpated on time-scales relevant for
management.
The substantial differences in survival and reproduction that
we estimated between small and large sagebrush individuals
drive observed transient dynamics (Fig. 2). High mortality
probability along with comparatively little reproduction likely
lead to years of population decline before mortality can be
compensated for by larger individuals with high fecundity and
survival (Fig. 3). The long period of low population densities
and cover we estimate indicate that monitoring performed 3–
5 years after seeding (current practice by federal agencies)
may not accurately reflect long-term outcomes. Our inferred
vital rates correspond well with existing field studies (see
Table S2 for summary). For example, one study in central
Utah found no mortality in sagebrush in their largest size
class (canopy volume >175 cm
2
) over a 3-year period in
lightly grazed sites, and survival probabilities decreased in
smaller plants (Owens & Norton 1990). In a second study,
seedling survival was consistently <5% over the first growing
season (Owens & Norton 1992).
Population modeling approaches provide new insight into
the demographic mechanisms that contribute to observed gra-
dients in resilience and recovery. In particular, sagebrush pop-
ulations at high elevations are typically much more resilient to
disturbance (Knutson et al. 2014). Our results suggest that
this pattern is likely due to higher population growth rates
and weaker effects of transient dynamics at high elevation
sites. High elevation sites have higher soil moisture (Figs. S17,
S21), which lead to faster growth rates among small individu-
als, allowing them to more quickly grow out of small sizes
classes with low survival probabilities and low fecundities.
Sensitivity analyses indicate that the ability of individuals to
rapidly grow out of the smallest size classes to larger sizes can
significantly increase near-term population growth (Fig. S5).
One of the more unexpected results we found is that many
populations were expected to decline (k
(s)
<1) even under
long-term, asymptotic conditions, particularly at low elevation
(Fig. 4). We hypothesise that a few mechanisms could account
for declining populations in sites that were presumably occu-
pied by sagebrush before disturbance. First, parameter esti-
mates for each site implicitly include the impact of other
exogenous changes after disturbance on vital rates. For exam-
ple, increases in invasive annual plants are widespread in
burned areas of the Great Basin, particularly at low elevation,
and likely reduce sagebrush population growth (Germino
et al. 2018). Indeed, we find that asymptotic and observed k
(s)
both decline with increasing Bromus tectorum cover
(Fig. S22). Adding to this, restoration seed sources used after
disturbance could be maladapted to a given location and
reduce demographic performance and population growth
(Germino et al. 2019). Alternatively, sagebrush, like many
dryland species, experiences infrequent recruitment pulses that
are in some cases decades apart (Maier et al. 2001). Because
time between recruitment events could surpass the length of
many of our datasets we may not have been able to capture
infrequent recruitment events that would increase population
growth. However, even if recruitment pulses are able to main-
tain populations in the long-term, many populations may be
rapidly driven to local extinction after seeding and before a
recruitment pulse can occur, thus leading to transformation.
Finally, climate change may have pushed many low elevation
sites that formerly supported viable sagebrush populations to
warmer and potentially drier conditions that are no longer
suitable (Sax et al. 2013). High survival probabilities of large
individuals could allow sagebrush to continue to occupy sites
long after climate change has made them unsuitable for popu-
lation growth. Given the strong negative impacts of transient
dynamics in establishing populations, hypothesis 1 and 3
could also help explain the obvious question: How did sage-
brush populations ever become established? Past cool, wet,
and invasive-free conditions at low elevation sites (similar to
conditions at higher elevation sites now) may have allowed
for large-scale colonisation by sagebrush, and once estab-
lished, these populations were able to remain stable or grow
despite warming and drying. However, warmer and drier con-
ditions coupled with increases in invasive species and the size
of wildfires may now make landscape-scale recovery impossi-
ble for many populations (Stevens-Rumann et al. 2018).
Quantifying the drivers of initial plant recruitment following
seeding has been a central focus of valuable restoration
research (Hardegree et al. 2018; Shriver et al. 2018). Specifi-
cally, research has identified how variable climate conditions
present a challenge for restoration of dryland ecosystems, but
targeting restoration seeding to favorable conditions (e.g.
above-average precipitation) can provide windows of opportu-
nity to increase seedling establishment (Holmgren & Scheffer
2001; Davis et al. 2016). Our results confirm key findings of
research focused on early recruitment conditions, specifically
our sensitivity analyses suggest the growth of small, newer
individuals to larger sizes (with higher survival) is an impor-
tant driver of population growth during recovery (Fig. S5).
Our results also highlight previously unquantified challenges
in restoration following early establishment. Although favor-
able early weather conditions are likely to increase population
density and decrease the likelihood of extirpation, altered
population structures and resulting transient population
dynamics may still lead to low density and restoration failure.
Even if a species existed and had stable populations prior to
disturbance, or other feedbacks (e.g. invasives) are eliminated,
restoration of stable populations may not be possible in the
near-term without intense management intervention. Still,
these insights provide opportunities to improve land manage-
ment. First, demographic analyses can provide a clear quan-
tification of where restoration is likely to be successful, results
that are important for managers who wish to allocate
Published 2019. This article is a U.S. Government work and is in the public domain in the USA
8Shriver et al. Letter
resources to areas that may benefit from restoration and avoid
areas where resources would be ineffectively used (Larios
et al., 2017). Second, demographic models can be powerful
tools to identify and test alternative restoration strategies. For
example, hand planting of large individuals may increase sur-
vival and provide consistent seed sources for new recruits
(Pyke 2011). At troublesome sites, these approaches could be
alternative or complementary techniques to the one-and-done
seedings often used (Young & Evans 2001). The efficacy of
alternative restoration strategies now and in a warmer more
variable future could be tested using our population modelling
approach.
Recognising the role of demographic processes in regulating
ecosystem responses to disturbance will be critical for accu-
rately anticipating the impact of global change. Disturbances,
natural and anthropogenic, are expected to be more widespread
in the next century. If populations are unable to effectively
recover from these disturbances, ecosystems may reach thresh-
olds where foundational species are lost and ecosystems rapidly
and unexpectedly transition into another state (Ratajczak et al.
2018). As our results indicate, even sites that support seemingly
stable populations may be slow to overcome the demographic
hurdles that drive transient population declines following larger
or severe disturbances. Importantly, the impact of transient
population dynamics on disturbance recovery in a given ecosys-
tem can be inferred by quantifying underlying vital rates prior
to disturbance. Emerging data sources paired with data-driven
population modelling approaches provide exciting opportuni-
ties to better understand the demographic processes controlling
vegetation patterns and dynamics across landscapes (Gurevitch
et al. 2016). For example, demographic models parameterised
in extant forests and shrub populations could be used to simu-
late disturbance and quantify resilience based on the likelihood
and length of recovery. Such results could be used by resource
managers to quantify ecological resilience to disturbance and
provide valuable early warning indicators that anticipate
ecosystem transformations.
ACKNOWLEDGEMENTS
We thank Dan Doak and Charles Yackulic for their thought-
ful feedback on this manuscript. We also thank Kiona Ogle
for feedback on early versions of the model, and Jim Clark
for IPM example code. Funding was provided by the Bureau
of Land Management, US Fish and Wildlife Service, US Geo-
logical Survey, the Great Basin Landscape Conservation
Cooperative and Joint Fire Science Program (Grant #16-1-03-
13). We thank Karen Prentice, Stephen Small, Todd Hopkins,
Steve Hanser, Carol Schuler and Martin Fitzpatrick for their
assistance planning and funding the project. We thank the
BLM Field Office personnel for contributing information
about individual post-fire treatments. Any use of trade, pro-
duct or firm names is for descriptive purposes only and does
not imply endorsement by the US Government.
AUTHOR CONTRIBUTIONS
DSP, MJG, MCD, DAP and JBB were the lead investigators
on the Great Basin portion of the project. RSA, DSP, MJG
and DAP led the collection of field data with help from
numerous field assistants. RSA and JLW managed the data
and metadata. CMA performed eco-hydrological simulations
with help from RKS and JBB. RKS performed analysis and
wrote the first draft with help from JBB. All authors con-
tributed to initial manuscript conception and editing.
DATA ACCESSIBILITY STATEMENT
Data and STAN code is available in USGS Science Base;
Data are available at: (Shriver and Bradford, 2019) https://doi.org/
10.5066/P944D1YU.
REFERENCES
Albrecht, M.A., Osazuwa-Peters, O.L., Maschinski, J., Bell, T.J., Bowles,
M.L., Brumback, W.E., et al. (2019). Effects of life history and
reproduction on recruitment time lags in reintroductions of rare plants.
Conservation Biology., 33(3), 601–611.
Bureau of Land Management. (2016). Budget Justifications FY 2017.
Caswell, H. (2001). Matrix Population Models: Construction, Analysis, and
Interpretation, 2nd edn. Sinauer Associates, Sunderland, MA.
Caughlin, T.T., Damschen, E.I., Haddad, N.M., Levey, D.J. & Warneke,
C. (2019). Landscape heterogeneity is key to forecasting outcomes of
plant reintroduction. Ecol. Appl., 29, 1–14.
Chambers, J.C., Pyke, D.A., Maestas, J.D., Pellant, M., Boyd, C.S.,
Campbell, S.B., et al. (2014). Using resistance and resilience concepts to
reduce impacts of invasive annual grasses and altered fire regimes on
the sagebrush ecosystem and greater sage-grouse. Gen. Tech. Rep.
RMRS-GTR-326. U.S. Dep. Agric. For. Serv. Rocky Mt. Res. Stn.1–
73.
Copeland, S.M., Munson, S.M., Pilliod, D.S., Welty, J.L., Bradford, J.B.
& Butterfield, B.J. (2018). Long-term trends in restoration and
associated land treatments in the southwestern United States. Restor.
Ecol., 26, 311–322.
Crone, E.E., Ellis, M.M., Morris, W.F., Stanley, A., Bell, T.,
Bierzychudek, P., et al. (2013). Ability of matrix models to explain the
past and predict the future of plant populations. Conserv. Biol., 27(5)
968–978.
Dalgleish, H.J., Koons, D.N., Hooten, M.B., Moffet, C.A. & Adler, P.B.
(2011). Climate influences the demography of three dominant sagebrush
steppe plants. Ecology, 92, 75–85.
Davies, K.W., Boyd, C.S., Beck, J.L., Bates, J.D., Svejcar, T.J. & Gregg,
M.A. (2011). Saving the sagebrush sea: An ecosystem conservation
plan for big sagebrush plant communities. Biol. Conserv., 144, 2573–
2584.
Davis, F.W., Sweet, L.C., Serra-diaz, J.M., Franklin, J., Mccullough, I.,
Flint, A., et al. (2016). Shrinking windows of opportunity for oak
seedling establishment in southern California mountains,7,1–18.
Doak, D.F. & Morris, W.F. (1999). Detecting population-level
consequences of ongoing environmental change without long-term
monitoring. Ecology, 80, 1537–1151.
Easterling, M.R., Ellner, S.P., Dixon, P.M. & Mar, N. (2000). Size-
specific sensitivity: Applying a new structured population model.
Ecology, 81, 694–708.
Germino, M.J., Barnard, D.M., Davidson, B.E., Arkle, R.A., Pilliod,
D.S., Fisk, M.R., et al. (2018). Thresholds and hotspots for shrub
restoration following a heterogeneous megafire. Landsc. Ecol., 33,
1177–1194.
Germino, M.J., Moser, A.M. & Sands, A.R. (2019). Adaptive variation,
including local adaptation, requires decades to become evident in
common gardens. Ecol. Appl., 29, 1–7.
Ghosh, S., Gelfand, A.E. & Clark, J.S. (2012). Inference for size
demography from point pattern data using integral projection models.
J. Agric. Biol. Environ. Stat., 17, 641–677.
Published 2019. This article is a U.S. Government work and is in the public domain in the USA
Letter Transient dynamics impede restoration 9
Gurevitch, J., Fox, G.A., Fowler, N.L. & Graham, C.H. (2016).
Landscape demography: Population change and its drivers across
spatial scales. Q. Rev. Biol., 91, 459–485.
Hardegree, S.P., Abatzoglou, J.T., Brunson, M.W., Germino, M.J.,
Hegewisch, K.C., Moffet, C.A., et al. (2018). Weather-centric rangeland
revegetation planning. Rangel. Ecol. Manag., 71, 1–11.
Hastings, A. (2016). Timescales and the management of ecological
systems. Proc. Natl Acad. Sci., 113, 14568–14573.
Hastings, A., Abbott, K.C., Cuddington, K., Francis, T., Gellner, G.,
Lai, Y., et al. (2018). Transient phenomena in ecology. Science, 361
(6406), eaat6412.
Holling, C.S. (1973). Resilience and stability of ecological systems. Annu.
Rev. Ecol. Syst.,4,1–23.
Holmgren, M. & Scheffer, M. (2001). El Nino as a window of
opportunity for the restoration of degraded arid ecosystems.
Ecosystems, 4, 151–159.
Hooten, M.B., Wikle, C.K., Dorazio, R.M. & Royle, J.A. (2007).
Hierarchical spatiotemporal matrix models for characterizingn
invasions. Biometrics, 63(2), 558–567.
Iles, D.T., Salguero-G
omez, R., Adler, P.B. & Koons, D.N. (2016).
Linking transient dynamics and life history to biological invasion
success. J. Ecol., 104, 399–408.
James, J.J., Svejcar, T.J. & Rinella, M.J. (2011). Demographic processes
limiting seedling recruitment in arid grassland restoration. J. Appl.
Ecol., 48, 961–969.
Johnson, E.A. & Miyanishi, K. (2008). Testing the assumptions of
chronosequences in succession. Ecol. Lett., 11, 419–431.
Jones, O.R., Scheuerlein, A., Salguero-G
omez, R., Camarda, C.G.,
Schaible, R., Casper, B.B., et al. (2014). Diversity of ageing across the
tree of life. Nature, 505, 169–73.
Kleinhesselink, A.R. & Adler, P.B. (2018). The response of big sagebrush
(Artemisia tridentata) to interannual climate variation changes across its
range. Ecology, 99, 1139–1149.
Knutson, K.C., Pyke, D.A., Wirth, T.A., Arkle, R.S., Pilliod, D.S.,
Brooks, M.L., et al. (2014). Long-term effects of seeding after wildfire
on vegetation in Great Basin shrubland ecosystems. J. Appl. Ecol., 51,
1414–1424.
Larios, L., Hallett, L.M. & Suding, K.N. (2017). Where and how to
restore in a changing world: a demographic-based assessment of
resilience, Journal of applied ecology, 54(4)1040–1050.
Loso, M.G. & Doak, D.F. (2006). The biology behind lichenometric
dating curves. Oecologia, 147, 223–9.
Maier, A.M., Perryman, B.L., Olson, R.A. & Hild, A.L. (2001). Climatic
influences on recruitment of 3 subspecies of Artemisia tridentata.J.
Range Manag., 54, 699–703.
Mcdonald, J.L., Stott, I., Townley, S. & Hodgson, D.J. (2016). Transients
drive the demographic dynamics of plant populations in variable
environments. J. Ecol., 104, 306–314.
Monnahan, C.C., Thorson, J.T. & Branch, T.A. (2017). Faster estimation
of Bayesian models in ecology using Hamiltonian Monte Carlo.
Methods Ecol. Evol., 8, 339–348.
Montero-Serra, I., Garrabou, J., Doak, D.F., Figuerola, L., Hereu, B.,
Ledoux, J.B., et al. (2018). Accounting for life-history strategies and
timescales in marine restoration. Conserv. Lett., 11, 1–9.
Nolan, C., Overpeck, J.T., Allen, J.R.M., Anderson, P.M., Betancourt,
J.L., Binney, H.A., et al. (2018). Past and future global transformation
of terrestrial ecosystems under climate change. Science, 361, 920–923.
Owens, M.K. & Norton, B.E. (1990). Survival of juvenile basin big
sagebrush under different grazing regimes. J. Range Manag., 43, 132–
135.
Owens, M.K. & Norton, B.E. (1992). Interactions of grazing and plant
protection on basin big sagebrush (Artemisia tridentata ssp. tridentata)
seedling survival. J. Range Manag., 45, 257–262.
Pilliod, D.S. & Welty, J.L. (2013). Land Treatment Digital Library. U.S,
Geol. Surv. Data Ser., p. 806.
Pilliod, D.S., Welty, J.L. & Toevs, G.R. (2017). Seventy-five years of
vegetation treatments on public rangelands in the Great Basin of North
America. Rangelands, 39, 1–9.
Pyke, D.A. (2011). Restoring and rehabilitating sagebrush habitats. Stud.
Avian Biol., 38, 531–548.
Ratajczak, Z., Carpenter, S.R., Ives, A.R., Kucharik, C.J., Ramiadantsoa,
T., Stegner, M.A., et al. (2018). Abrupt change in ecological systems:
Inference and diagnosis. Trends Ecol. Evol., 33, 513–526.
Richardson, B.A., Ortiz, H.G., Carlson, S.L., Jaeger, D.M., Shaw, N.L.
& Peters, D.P.C. (2015). Genetic and environmental effects on seed
weight in subspecies of big sagebrush: Applications for restoration.
Ecosphere,6,1–13.
Sax, D.F., Early, R. & Bellemare, J. (2013). Niche syndromes, species
extinction risks, and management under climate change. Trends Ecol.
Evol., 28, 517–523.
Scheffer, M., Carpenter, S., Foley, J.A., Folke, C. & Walker, B. (2001).
Catastrophic shifts in ecosystems. Nature, 413, 591–596.
Schlaepfer, D.R., Lauenroth, W.K. & Bradford, J.B. (2014). Natural
regeneration processes in big sagebrush (Artemisia tridentata). Rangel.
Ecol. Manag., 67, 344–357.
Shriver, R.K. & Bradford, J.B. (2019). Demographic modeling data
(including code) at various sites in the Great Basin. Geological Survey
data release USA. https://doi.org/10.5066/P944D1YU.
Shriver, R.K., Cutler, K. & Doak, D.F. (2012). Comparative demography
of an epiphytic lichen: support for general life history patterns and
solutions to common problems in demographic parameter estimation,
Oecologia, 170, 137–146.
Shriver, R.K., Andrews, C.M., Pilliod, D.S., Arkle, R.S., Welty, J.L.,
Germino, M.J., et al. (2018). Adapting management to a changing
world: Warm temperatures, dry soil, and inter-annual variability limit
restoration success of a dominant woody shrub in temperate drylands.
Glob. Chang. Biol., 24, 4972–4982.
Stevens-Rumann, C.S., Kemp, K.B., Higuera, P.E., Harvey, B.J., Rother,
M.T., Donato, D.C., et al. (2018). Evidence for declining forest
resilience to wildfires under climate change. Ecol. Lett., 21, 243–252.
Suding, K.N. & Hobbs, R.J. (2009). Threshold models in restoration and
conservation: a developing framework. Trends Ecol. Evol., 24, 271–279.
Suding, K.N., Gross, K.L. & Houseman, G.R. (2004). Alternative states and
positive feedbacks in restoration ecology. Trends Ecol. Evol., 19, 46–53.
Tredennick, A.T., Hooten, M.B., Aldridge, C.L., Homer, C.G.,
Kleinhesselink, A.R. & Adler, P.B. (2016). Forecasting climate change
impacts on plant populations over large spatial extents. Ecosphere,7,1–16.
West, N.E. (2000). Synecology and disturbance regimes of sagebrush
steppe ecosystems. In: ID Boise,Proceedings: sagebrush steppe
ecosystems symposium. USDI Bureau of Land Management Publication
BLM/ID/PT-00100111150, pp. 15–26.
Wijayratne, U.C. & Pyke, D.A. (2012). Burial increases seed longevity of two
Artemisia tridentata (Asteraceae) subspecies. Am. J. Bot., 99, 438–447.
Young, T.P. & Evans, R.Y. (2001). Container stock versus direct seeding
for woody species in restoration sites. Comb. Proc. Int. Plant
Propagators’ Soc., 50, 577–582.
SUPPORTING INFORMATION
Additional supporting information may be found online in
the Supporting Information section at the end of the article.
Editor, Shaopeng Wang
Manuscript received 14 November 2018
First decision made 22 December 2018
Second decision made 3 April 2019
Manuscript accepted 9 May 2019
Published 2019. This article is a U.S. Government work and is in the public domain in the USA
10 Shriver et al. Letter
A preview of this full-text is provided by Wiley.
Content available from Ecology Letters
This content is subject to copyright. Terms and conditions apply.
