Perspectives in Plant Ecology, Evolution and Systematics 15 (2013) 328–337
Contents lists available at ScienceDirect
Perspectives in Plant Ecology, Evolution and Systematics
jo urnal homepage: www.elsevier.com/locate/ppees
Research article
Conservatism of responses to environmental change is rare under
natural conditions in a native grassland
∗
Jonathan A. Bennett , James F. Cahill Jr.
Department of Biological Sciences, University of Alberta, Edmonton, AB T6G 2E9, Canada
a r t i c l e i n f o a b s t r a c t
Article history: Whether or not niche conservatism is common is widely debated. Despite this uncertainty, closely related
Received 27 April 2013
species are often assumed to be ecologically similar. This principle has led to the proposed use of phyloge-
Received in revised form
netic information in forecasting species responses to environmental change. Tests of niche conservatism
23 September 2013
often focus on ‘functional traits’ and environmental tolerances, but there have been limited tests for con-
Accepted 16 October 2013
servatism in species’ responses to changes in the environment, especially in the field. The prevalence
Available online 24 October 2013
of functional convergence and the likelihood of functional trade-offs in a heterogeneous environment
suggest that conservatism of the response niche is unlikely to be detectable under natural conditions. To
Keywords:
test the relevance of evolutionary information in predicting ecological responses, we tested for conser-
Phylogenetic community ecology
vatism (measured as phylogenetic signal) of grassland plant population responses to 14 treatments (e.g.
Phylogenetic signal
Phylogenetic niche conservatism light, nutrients, water, enemies, mutualists), each manipulated for 2–3 years, and 4 treatment categories
Belowground ecology (aboveground, belowground, resource, and herbivory) at a single site. Individual treatment responses
Grazing showed limited evidence of conservatism, with only weak conservatism in plant responses to mycorrhi-
Mycorrhizae
zae and grazing. Aspects of the response niche were conserved among monocots both aboveground and
belowground, although the pattern varied. Conservatism was limited to grazing aboveground, but below-
ground responses were conserved as a group, suggesting fundamental differences in how selection has
led to niche conservatism in aboveground and belowground environments. Overall, our results suggest
that conservatism of the response niche is not common, but is actually rare. As such, evolutionary rela-
tionships are likely to be of limited relevance for predicting species responses under field conditions, at
least over the short time scales used in this study.
© 2013 Elsevier GmbH. All rights reserved.
Introduction many ecological factors differentially affect certain lineages within
the community, causing phylogenetic clustering (Helmus et al.,
Plant populations often respond idiosyncratically to changes in 2010; Verdú and Pausas, 2007). This suggests that phylogeny can be
their environment (Tilman, 1987; Turkington et al., 2002). Efforts used as a tool to predict species responses to changes in their envi-
have been made to identify species characteristics that can be ronment, but for phylogeny to be a useful predictor of ecological
used to develop a predictive framework for changes in the relative responses, the niche must be conserved. However, the prevalence
abundance of plant populations (e.g. Grime, 1977; Westoby, 1998). of niche conservatism has been questioned (Knouft et al., 2006;
Based upon the idea that related species are more ecologically Lavergne et al., 2010; Losos, 2008; Silvertown et al., 2006b).
similar (Darwin, 1859), hypothesized patterns of descent (e.g. a Niche conservatism can have multiple definitions. Here, we
phylogeny) have been used with some success in determining how define niche conservatism broadly as the tendency of related
species respond to both biotic (Burns and Strauss, 2011; Reinhart species to respond similarly to abiotic or biotic environmental
et al., 2012) and abiotic (Niinemets and Valladares, 2006; Prinzing, conditions (Wiens et al., 2010; Wiens and Graham, 2005). This def-
2001; Willis et al., 2008) elements of their environments. Further, inition is more liberal than other definitions that consider niche
conservatism to require species being more similar than expected
under a model of Brownian evolution (Losos, 2008). While phy-
logenetic relatedness is often considered an integrative measure
∗
Corresponding author at: B715 Biological Sciences Building, University of of functional similarity (Mouquet et al., 2012; Webb et al., 2002),
Alberta, Edmonton, AB T6G 2E9, Canada. Tel.: +1 780 492 1577;
for plants, ecologically relevant traits are often labile (Cavender-
fax: +1 780 492 9234.
Bares et al., 2006; Grime, 2006) or environmentally plastic (Berg
E-mail address: [email protected] (J.A. Bennett).
1433-8319/$ – see front matter © 2013 Elsevier GmbH. All rights reserved. http://dx.doi.org/10.1016/j.ppees.2013.10.001
J.A. Bennett, J.F. Cahill Jr. / Perspectives in Plant Ecology, Evolution and Systematics 15 (2013) 328–337 329
and Ellers, 2010; Burns and Strauss, 2012). Further, there are many tested for niche conservatism (as measured by phylogenetic sig-
ways to respond to different aspects of the environment. For exam- nal) in responses to each individual treatment and in responses
ple, defensive compounds are produced using different pathways, to four categories of ecological treatments representing resource,
but all reduce herbivory (Howe and Jander, 2008) and competitive herbivory, aboveground, and belowground treatment groupings.
response is associated with many traits representing different ways
of coping with reduced resource availability (Wang et al., 2010).
Materials and methods
Additionally, traits may be associated with multiple functions, yet
multiple traits may determine a species’ functional response to a Site description
given factor. High volumes of fine roots can increase both nitrogen
and water uptake (Craine et al., 2003), but periodic drought tol- All experiments occurred at the approximately 5000 ha Univer-
◦
erance also requires that the plant be able to store water for later sity of Alberta research ranch at Kinsella, Alberta, Canada (53 05 N,
◦
use, which is not an adaptation related to nitrogen uptake (Craine, 111 33 W). Research occurred in three fields located in two sep-
2009). This suggests that conservatism of a trait does not mean arate sections of the ranch totalling 100 ha. Field 1 was located
that a plant’s response to one factor related to that trait can predict in the northern part of the ranch, whereas fields 2 and 3 are in
its response to other related factors. Many of the traits necessary the southern part of the ranch, with the two sites separated by
to respond to environmental conditions also involve functional approximately 6 km. The fields used are unseeded and unbroken
trade-offs, such as those between shade and drought tolerance and represent a savannah habitat with mixed grass prairie (pri-
(Niinemets and Valladares, 2006). As a consequence, plant species marily Hesperostipa curtiseta (Hitchc.) Barkworth, Poa pratensis L.
may be suited to cope with certain environmental conditions, but and Festuca hallii (Vasey) Piper) interspersed with stands of aspen
not others. Thus, for many reasons, ecological responses are often (Populus tremuloides Michx.). Though historically lightly grazed by
less conserved than morphological or physiological traits (Losos, cattle, grazing was halted for the duration of each experiment.
2008; Prinzing, 2001). This suggests that evolutionary informa- As is true for many grasslands (Foster et al., 2004; Silvertown
tion may be of limited use for predicting how species respond to et al., 2006c; Tilman, 1996), plant community structure and func-
environmental conditions in nature. tion varies spatially and temporally. Soils at the site have a thin
When suites of traits appear to confer specific functioning, they topsoil layer over glacial till (Lamb, 2008), but are spatially vari-
have often been grouped into plant functional strategies (Reich able in texture, chemistry, and topography (Bennett et al., 2013).
◦
et al., 2003; Westoby, 1998). Most commonly, plant strategies are The site has an average annual temperature of 2.8 C and receives
associated with responses to resource availability and disturbance, approximately 430 mm of precipitation in an average year, but is
where some species are adapted to quick growth and rapid resource subject to periodic drought (Cahill, 2003). Fig. 1 shows the vari-
acquisition, while others are adapted to disturbances such as her- ability in species richness, productivity, and phylogenetic diversity
bivory (Craine, 2009; Grime, 1977; Reich et al., 2003). Responses to over the duration of the experiments. These data were taken from
both resources and herbivory are often consistent within broad, un-manipulated plots at the field site, with species richness and
2
phylogenetically distinct functional groups (Coughenour, 1985; phylogenetic diversity data derived from cover estimates (0.25 m )
Lavorel et al., 1997; Niinemets and Valladares, 2006; Turkington and productivity estimates from live biomass clipped in small plots
2
et al., 2002), yet the evidence for conservatism of traits represent- (0.10 m ), dried, and weighed. Phylogenetic diversity was calcu-
ing these plant strategies is mixed (Brunbjerg et al., 2012; Diaz lated using the constructed phylogeny (see below) as abundance
et al., 2004). While there are a few experimental tests for conser- weighted mean phylogenetic distance (Webb et al., 2002) using the
vatism of plant strategies, to our knowledge, no studies have tested independent swap null model (Gotelli, 2000) in the picante package
whether population responses to multiple treatments related to in R (Kembel et al., 2010). Species richness and productivity were
these strategies are conserved. variable among years and across fields, with field 1 having higher
Plant strategies require coordinated responses to multiple envi- species richness and productivity than fields 2 and 3. Phyloge-
ronmental factors, both above- and belowground. This requires netic diversity was much more consistent across fields, with limited
that root and shoot traits co-vary. There is evidence for such interannual variation (Fig. 1), suggesting greater consistency in the
covariance (Craine et al., 2001, 2002), although root and shoot phylogenetic structure of the community.
traits may have evolved independently (Kembel and Cahill, 2011).
Individual root and shoot traits show varying degrees of conser-
Data selection
vatism (Anderson et al., 2011; Comas et al., 2012; Diaz et al.,
2004; Grime and Mackey, 2002; Kembel and Cahill, 2005, 2011), Data were taken from six separate multi-year multi-factorial
as do plant responses to various above- and belowground factors experiments, containing a total of 14 treatments (Table 1). Within
(Niinemets and Valladares, 2006; Prinzing, 2001; Silvertown et al., each independent experiment, interactions among the treatments
2006b). However, it is unclear whether plant responses to either were included in the original study. However, we do not have data
aboveground or belowground factors as groups would be phylo- testing the interactive effects of all treatments, so we limit our anal-
genetically conserved and there are no experimental tests of this yses to main effects, though we recognize complex interactions
concept. among this number of treatments can occur. For each treatment,
To test whether related species responded similarly to changes we only used plots where a single treatment was applied and
in their environment, and thus if response niches were conserved, compared those treatment plots to a control plot with no treat-
we synthesized the results of six short-term (2–3 years) experi- ments applied within the same block. Grazing was simulated by
ments conducted in a single grassland system within the Aspen clipping plants at either low intensity (7 cm stubble height) or
Parkland eco-region of Canada. In total, 14 abiotic and biotic treat- high intensity (3 cm stubble height; White et al., 2012). Fertilizers
2
ments were manipulated: aboveground insecticide; belowground were applied either as ammonium nitrate (5.4 g N/m ) for nitro-
insecticide; contact fungicide; drought; fixed interval watering; gen only (Lamb, 2008) or as a slow-release fertilizer for NPK at
2 ®
high intensity clipping; litter removal; low intensity clipping; 5.2 g NPK/m (applied as 14:14:14 Osmocote Classic, Scotts, Ben-
nitrogen addition; nitrogen, phosphorus and potassium (NPK) nett et al., in preparation). Fixed interval watering increased total
addition; shading; systemic fungicide; variable interval watering; precipitation by 50% through weekly water addition (Lamb, 2008).
and warming. From population responses to these treatments, we Drought treatments decreased precipitation by 60% using rainout
330 J.A. Bennett, J.F. Cahill Jr. / Perspectives in Plant Ecology, Evolution and Systematics 15 (2013) 328–337
decrease nutrient availability (Sardar and Kole, 2005) and can
be harmful to other organisms (vandenBrink et al., 1996). Beno-
myl is commonly used to suppress arbuscular mycorrhizal fungi
(e.g. Hartnett and Wilson, 1999), but it also has effects on non-
target organisms and can increase available nitrogen (Allison et al.,
®
2007). Rovral has also been used to suppress mycorrhizal fungi
(Gange et al., 1990), but has fewer documented effects on non-
target organisms (Ganade and Brown, 1997) and no detectable
effects on soil nutrient availability (J.A. Bennett, unpublished data).
Detailed methods for each treatment can be found in the origi-
nal manuscripts. Additional methods details for the unpublished
experiment are found in Table 1.
Relative abundance was estimated as percent vegetative cover,
a commonly used method to assess relative change within herba-
ceous plant communities (Lamb and Cahill, 2008; Tilman, 1987).
Most experiments were 2–3 years long, but some experiments ran
for longer than 3 years and only percent cover estimates were col-
lected in the interim as destructive harvesting was unfeasible. To
minimize the variation in experimental duration, we chose to limit
experimental duration to 3 years. Across all experiments, we calcu-
lated the responses of 54 different species to at least one treatment,
with an additional six species included when calculating aggregate
responses to treatment categories (see below). For most species,
we were able to calculate responses to approximately half of the 14
treatments (mean 7.1, standard deviation 4.19), although species
did vary in how often we were able to calculate responses (see
Figure A1). Our abundance estimates were the mean of three cover
estimates taken over the growing season (late spring, mid-summer,
2
and late summer) from 0.25 m sub-plots within each larger con-
trol or treatment plot. Changes in relative abundance for each
species were calculated as the log response ratio of abundances
(ln(treatment/control)) for each pair of treatment and control plots.
The log response ratio was used instead of percent change to nor-
malize responses (Hedges et al., 1999). As many experiments were
multi-factorial, control plots within a given block were used in
the calculation of responses to multiple treatments. For exam-
ple, when calculating responses to aboveground and belowground
insect suppression, the control plot where no treatments were
applied was used to calculate responses to both aboveground insect
suppression and belowground insect suppression. However, plots
where insects were suppressed above- and belowground were not
Fig. 1. Inter-annual variability in (B) species richness, (C) phylogenetic diversity,
included in any calculations. From these measurements, we calcu-
calculated as abundance weighted mean phylogenetic distance, and (D) standing
lated the average response of a species to each treatment and the
biomass over the course of the 14 experimental manipulations. Panel A shows the
standard error of that estimate. By using the average response of
time frame over which each of the manipulations was conducted. In panels B–D,
empty circles show conditions in field 1, grey circles in field 2, and black circles a species, we ignore the potential variation in how individuals of
in field 3. Experimental manipulations abbreviated as follows: AI – aboveground a given species respond to a treatment due to neighbour compo-
insecticide, BI – belowground insecticide, CF – contact fungicide, D – drought, FW –
sition or local environmental conditions. Given that we are testing
fixed interval watering, HC – high intensity clipping, LC – low intensity clipping, LR
the generality of niche conservatism and the utility of evolutionary
– litter removal, N – nitrogen addition, NPK – NPK addition, S – shade, SF – systemic
fungicide, VW – variable interval watering, and W – warming. Error bars in panels information for informing ecology under natural conditions, the
B–D represent standard error. particular set of conditions under which related species respond
similarly is not of interest in the current study.
For each treatment, the mean change in abundance across
shelters (White et al., 2012). This water was collected and added to populations and its 95% confidence intervals were estimated
plots within 24 h following rainfall for the variable interval water- using the average response of each species to each treatment in a
ing treatment (White et al., 2012). Shade cloth was used to reduce mixed model in SPSS (v. 19.0). We only included species for which
light by 73% (Lamb, 2008), open-top chambers were used to warm we could calculate the standard error for a given treatment and
◦
plots by approximately 3 C (White et al., 2012), and litter was weighted species responses by the inverse of that standard error.
removed by raking (Bennett et al., in preparation). Insects were We also attempted to account for additional sources of variation by
TM
suppressed using chlorpyrifos with Lorsban 4E (Dow) used for including a number of random factors, including year of data col-
TM
aboveground insects and Lorsban 5G (Dow) for belowground lection and experimental duration. As experiments at the site were
insects (Clark et al., 2012; Coupe et al., 2009). Fungi were sup- spatially distinct, we included experimental identity to account for
pressed using both a systemic fungicide (Benomyl, Dupont Inc.; spatial variability among and within fields, but nested it within year
®
Cahill et al., 2008a) and a contact fungicide (Rovral , Bayer; Ben- as data from multiple experiments was collected in most years.
nett et al., in preparation). As with all pesticides, each of the Species identity was also included as a random factor to account
pesticides used has non-target effects. Chlorpyrifos is known to for differences in species pool across factors. Thus, we included
J.A. Bennett, J.F. Cahill Jr. / Perspectives in Plant Ecology, Evolution and Systematics 15 (2013) 328–337 331
Table 1
Meta data for each treatment included in the analysis.
a b c d e
Treatment Category Harvested Length (years) Field Species N Dir. Methods
Aboveground insecticide A,H 2003, 2005 2, 3 1, 2 48 22 + Clark et al. (2012) and Coupe et al. (2009)
Belowground insecticide B,H 2005 3 1 41 10 + Coupe et al. (2009)
f
Contact fungicide U,U 2010 3 1 41 20 ? Bennett et al. (in preparation)
Drought B,R 2010 3 3 10 5 − White et al. (2012)
Fixed watering B,R 2005 3 1 46 22 + Lamb (2008)
High clipping A,H 2010 3 3 15 5 − White et al. (2012)
f
Litter A,U 2010 2 1 41 20 ? Bennett et al. (in preparation)
Low clipping A,H 2010 3 3 15 5 − White et al. (2012)
Nitrogen B,R 2005 3 1 45 22 + Lamb (2008)
f
NPK B,R 2010 2 1 41 20 + Bennett et al. (in preparation)
−
Shading U,U 2005 3 1 45 22 Lamb (2008)
Systemic fungicide U,U 2005 3 1 38 20 ? Cahill et al. (2008a)
Variable watering B,R 2010 3 3 12 5 + White et al. (2012)
Warming U,U 2010 3 3 16 5 ? White et al. (2012)
a
High and low refer to the intensity of clipping; above and below refer to aboveground and belowground; fixed and variable refer to the interval of watering.
b
Treatments are classified as aboveground (A) or belowground (B) and herbivory (H) or resource-based (R). Treatments we could not classify are categorized as unknown (U).
c
Species refers to the number of species for which we could calculate the standard error of their response to that treatment, allowing us to estimate their response to that treatment.
d
N refers to the number of paired treatment and control plots in the experiment where the treatment was applied.
e
Treatments were classified as having a positive (+), negative (−) or unknown (?) hypothesized direction of effect.
f ® 2
Rovral (Bayer) was applied to half the plots at a rate of 0.36 g/m active ingredient (iprodione) every two weeks. Litter was raked each spring in all plots, replaced in
®
control plots and disposed of in litter removal plots. Fertilizer was added as 3- to 4-month slow release 14:14:14 nutrient pellets (Osmocote , Scotts) each spring at a rate
2
of 5.22 g NPK/m .
treatment as a fixed effect and the calendar year the data was present except amongst close relatives within Poaceae and Aster-
collected, the duration of the experiment, experiment identity aceae (Fig. A1). Polytomies at the tips of a phylogenetic tree are
nested within year and species identity as random effects in the unlikely to influence analyses of phylogenetic signal (Münkemüller
model. In the final model, we retained only species identity among et al., 2012; Swenson, 2009).
the random effects as the other random effects explained no
additional variation, resulting in a Hessian matrix that was not Niche conservatism
positive definite.
Given that there is spatial and temporal heterogeneity in local
Our definition of niche conservatism – related species respond
community processes (Fig. 1), we explored the amount of variation
similarly to ecological factors – is broad and our approach is holis-
explained by year and experimental identity relative to the treat-
tic in its focus on population outcomes, rather than trait-focused
ments to evaluate our decision to exclude them from the model. We
measures of plant morphology or physiology. We used three sep-
ran a mixed model with year, experimental identity nested within
arate methods to test for phylogenetic signal as a proxy for niche
year, and treatment identity nested within experimental identity
conservatism: Blomberg’s K (Blomberg et al., 2003), Pagel’s (Pagel,
within year as fixed factors. As in the previous model, species
1999), and the decomposition of trait variation (Pavoine et al.,
identity was included as a random effect and species responses
2010). The first two methods assess whether the distribution of
were weighted by the inverse of the standard error. Only treat-
traits (or in this case population responses) among species fol-
ment identity (F = 2.73, P = 0.004) and not experiment identity
9,421 lows a Brownian motion model of evolution. The third method
(F = 1.34, P = 0.261) or year (F = 1.45, P = 0.236) explained
3,450 2,456 assesses response diversity among all the species descending from
significant variation in species responses, supporting our decision
each branch of each node of the phylogenetic tree, measured as
to remove these random factors. We recognize that this does not
quadratic entropy (Rao, 1982). This information is used to gen-
account for potential differences in how species may respond to
erate both a visual display of where divergence occurred along
treatments in different years, but given the nature of the data such
the tree and uses randomization procedures to determine if there
a test is not feasible.
is significant phylogenetic signal. The randomization tests for the
response decomposition analyses indicate whether response vari-
Phylogenetic information ation is skewed towards a single node, a few nodes, the root, or the
tips (Pavoine et al., 2010). Variation skewed towards the root or
Phylogenetic information was extracted from the molecu- towards one or few nodes can be used to infer niche conservatism or
lar phylogeny outlined in Bennett et al. (2013) that sampled at least differentiation, whereas variation skewed towards the tips
146 species across 35 families found at the study site (see suggests convergence. However, careful examination of how vari-
Fig. A2). The phylogeny was based on a 1400 bp section of the ation in responses is distributed across the nodes of the phylogeny
ribulose-biphosphate carboxylase gene (rbcL) and constructed is required to infer patterns of niche conservatism.
using standard techniques. Although the phylogeny only sampled Before quantifying phylogenetic signal, we created separate
one gene, sequence variation in rbcL was sufficient to resolve rela- phylogenetic trees for each experimental treatment, for a total of
tionships with strong support. Deeper branching patterns were 14 trees. Each tree was created by pruning the full phylogenetic
consistent with published angiosperm phylogenies based on mul- tree to include only species for which we had a response value
tiple genes (Bremer et al., 2009; Soltis et al., 2011) and topology with an associated error measurement in that treatment. We cal-