Ecological Applications, 20(7), 2010, pp. 1903–1912
Ó 2010 by the Ecological Society of America
Reconciling contradictory findings of herbivore impacts on spotted
knapweed (Centaurea stoebe) growth and reproduction
DAVID G. KNOCHEL1
AND
TIMOTHY R. SEASTEDT
Department of Ecology and Evolutionary Biology, Institute of Arctic and Alpine Research, University of Colorado,
Boulder, Colorado 80309 USA
Abstract. Substantial controversy surrounds the efficacy of biological control insects to
reduce densities of Centaurea stoebe, a widespread, aggressive invasive plant in North
America. We developed a graphical model to conceptualize the conditions required to explain
the current contradictory findings, and then employed a series of manipulations to evaluate C.
stoebe responses to herbivores. We manipulated soil nitrogen and competition in a field
population and measured attack rates of a foliage and seed feeder (Larinus minutus), two gall
flies (Urophora spp.), and a root feeder (Cyphocleonus achates), as well as their effects on the
growth and reproduction of C. stoebe. Nitrogen limitation and competing vegetation greatly
reduced C. stoebe growth. L. minutus most intensively reduced seed production in lownitrogen soils, and removal of neighboring vegetation increased Larinus numbers per flower
head and the percentage of flowers attacked by 15% and 11%, respectively. Cyphocleonus
reduced flower production and aboveground biomass over two years, regardless of resources
or competition. Our results, in conjunction with other published studies, demonstrate that
positive, neutral, and negative plant growth responses to herbivory can be generated.
However, under realistic field conditions and in the presence of multiple herbivores, our work
repudiates earlier studies that indicate insect herbivores increase C. stoebe dominance.
Key words: biological control; Centaurea stoebe; compensation; Cyphocleonus achates; herbivory;
Larinus minutus; plant competition; resource limitation; spotted knapweed; Urophora spp.; weed
management.
INTRODUCTION
Spotted knapweed (Centaurea stoebe L. subsp. micranthos [Gugler] Hayek [Asteraceae]), also identified as
C. maculosa and C. biebersteinii (see Ochsmann 2001,
Hufbauer and Sforza 2008), is one of several species of
the Eurasian knapweeds (Centaurea spp.) that have
become dominant components of the rangeland vegetation in western North America (Sheley et al. 1999). C.
stoebe now occupies over three million ha in rangeland
and forest ecosystems in North America, and 13 species
of insects have been released since 1970 in an attempt to
reduce its abundance (Story and Piper 2001). However,
the effectiveness of insect introduction as a sustainable
method to control the plant remains controversial, and
other unsustainable and cost-prohibitive management
techniques still predominate. Further, despite intense
scientific inquiry aimed at understanding and explaining
the unusually high dominance seen in this invasive
species, we lack information on how reintroduced
natural enemies, resource availability, and plant competition might collectively influence densities and
population dynamics of the plant.
Manuscript received 20 October 2009; revised 15 December
2009; accepted 16 December 2009. Corresponding Editor: T. J.
Stohlgren.
1 E-mail: [email protected]
The biological control insects studied here, a flower
head weevil Larinus minutus Gyllenhal (Coleoptera:
Curculionidae), two gall fly species of the genus
Urophora, U. affinis Frfld and U. quadrifasciata
Meigen (Diptera: Tephrididae), and the root-feeding
weevil, Cyphocleonus achates Fahr. (Coleoptera:
Curculionidae), have been at least locally abundant at
sites across North America for about two decades.
These insects established on spotted knapweed, but to
date, reports of significant population reductions
attributed to insect effects are few (Corn et al. 2006,
Cortilet and Northrop 2006, Jacobs et al. 2006, Story et
al. 2006, 2008, Michels et al. 2009). The perennial
growth habit of spotted knapweed likely improves its
capacity to maintain dense populations, despite reductions of seed production and tissue damage from
biological control insects.
Other field studies of a root-feeding moth (Agapeta
zoegana L. [Lepidoptera: Tortricidae]) and clipping
treatments to simulate herbivory demonstrated that
plant fitness may improve if plants respond to damage
by releasing allelopathic compounds that harm adjacent
vegetation (Callaway et al. 1999, Thelen et al. 2005,
Newingham et al. 2007). The efficacy of these compounds has been challenged (e.g., Chobot et al. 2009,
Duke et al. 2009); however, competitive outcomes in
response to herbivory may be valid, regardless of the
specific mechanism(s) producing this response. Urophora
1903
1904
DAVID G. KNOCHEL AND TIMOTHY R. SEASTEDT
FIG. 1. A hypothetical gradient of spotted knapweed
(Centaurea stoebe) fitness responses to herbivory, as regulated
by the combined influence of resource availability and plant
competition. This is analogous to the compensatory continuum
model proposed by Maschinski and Whitham (1989), except
that here the intensity rather than the timing of herbivory is
considered. At a given intensity of herbivory (sliding the vertical
bar left or right) spotted knapweed could potentially (A)
overcompensate, (B) equally compensate, or (C) undercompensate for damage, depending on the levels of resource
availability. Further, the presence of competition from neighboring vegetation could shift the response downward at a given
resource level.
gall flies may also indirectly benefit spotted knapweed
through interactions with deer mice that reduce seed
abundance of native competitors (Pearson and Callaway
2008). Further, others hypothesized that a persistent
drought throughout the western United States contributed at least partially to population declines (Corn et al.
2007, Pearson and Fletcher 2008). Thus the extent to
which these insects might control C. stoebe over the
three million ha of invaded rangelands remains uncertain. Similar to the theoretical approach used by Shea et
al. (2005) to assess biological control impacts on an
invasive thistle, we suggest that variation in resource
conditions and plant competition are likely to govern
spotted knapweed responses to herbivory and ultimately
govern population growth. A complete analysis of how
environmental drivers interact with herbivory is essential
in order to understand or predict the long-term
relationship between C. stoebe populations and specialist biological control insects.
Here we propose an herbivory–resource gradient
model, which describes how resource constraints and
plant competition moderate the interactions between C.
stoebe individuals and biological control insects (Fig. 1).
The model is built upon the conceptual frameworks of
McNaughton (1979), Maschinski and Whitham (1989),
and most recently, Wise and Abrahamson (2007). In the
model, a range of plant responses to herbivory may
occur, from negative (undercompensation), to zero
(equal compensation), to positive (overcompensation).
Likewise, the intensity of damage due to herbivory may
Ecological Applications
Vol. 20, No. 7
range from low to high, depending on both the identities
and densities of biological control insects present and
whether multiple species are feeding on root and foliar
tissues and seeds (Knochel 2009). In this context, a
range of local resource conditions (e.g., soil nutrients,
water, sunlight) and competition for these resources by
adjacent vegetation will determine how C. stoebe plants
fare. Collectively, these plant–insect–resource interactions at small spatial scales are hypothesized to govern
spotted knapweed dominance and its population dynamics. These interactions may explain the high
variability in the previously discussed outcomes observed for biological control insects on C. stoebe.
In a field population of C. stoebe, we measured the
effects of herbivory by specialist root-feeding
(Cyphocleonus achates) and stem-, foliage-, and flowerhead-feeding (Larinus minutus, Urophora spp.) biological control insects, which were allowed free access to
plants in plots that differed in levels of soil nitrogen (N)
availability and plant competition. We tested three
predictions related to the model shown in Fig. 1. (1)
Does spotted knapweed lose its competitive advantage
when soil resources are in low supply, and is the plant
susceptible to competition from neighboring plants? (2)
Does a gradient in soil nutrient availability and
neighboring plant cover affect attack rates of biological
control insects? (3) Do the effects on plant performance
by root-feeding and flower head insects depend on N
supply or neighboring plant competition? This field
experiment sought to demonstrate the range of potential
responses that occur when spotted knapweed interacts
with variable levels of soil resources, plant competitors,
and biological control insects.
METHODS
Field site description
Field research was conducted 15 km northwest of
the city of Boulder, Colorado, USA (40807 0 1400 N,
105819 0 2600 W), between 1865 m and 2070 m elevation,
on meadows and on rugged 15–60% slopes that are
dominated by ponderosa pine (Pinus ponderosa Douglas
Ex. Lawson). Soils at the site are composed of the Fern
Cliff-Allens Park-Rock outcrop complex, and the soils
of experimental meadows within the stream drainage are
of mixed loamy alluvium parent material, with a stony,
sandy clay loam profile at 0–100 cm depth (National
Resources Conservation Service 2008). Mean annual air
temperature is 6–88C, with a frost-free period of 80120
days. In the mid 1980s, Centaurea stoebe was accidentally introduced to this area and has spread over 40 ha of
meadows, riparian areas, and forest that burned in a
wildfire in 1988. This fire apparently facilitated the
invasion of other nonnative forbs, including dalmatian
toadflax (Linaria dalmatica [L.] Mill.), sulfur cinquefoil
(Potentilla recta L.), and musk thistle (Carduus nutans
L.) Meadows also contain nonnative crested wheatgrass
(Agropyron cristatum) and the invasive cheatgrass
October 2010
CONTEXT-DEPENDENT CONTROL OF C. STOEBE
1905
(Bromus tectorum), as well as relics of the mixed-grass
prairie, including blue grama, (Bouteloua gracilis [HBK]
Lag. ex Steud.), sideoats grama (Bouteloua curtipendula
[Michx.] Torr.), needle grass (Hesperostipa comata [Trin.
& Rupr.] Barkworth), and beebalm (Monarda fistulosa).
A history of moderate cattle grazing within the study
area occurred since the mid 1900s, and experiments were
fenced to prevent grazing by cattle or native ungulates.
Biological control insects were first observed at the site
in 2001, including two gall flies of the Urophora genus,
U. affinis and U. quadrifasciata, and the seed-head
weevil Larinus minutus. The seed-head insects were
supplemented from 2001–2005, prior to the start of the
experiment in 2007, with releases of approximately 3000
Larinus minutus weevils and 2000 Cyphocleonus achates
root weevils.
and carbon (C) addition (as sucrose) to reduce N
availability (low N). To increase soil N availability, we
added N as granular NO3NH4 at a rate of 38 g
Nm2yr1 in three separate additions on 2 May, 5
June, and 9 July in both 2007 and 2008. Ambient levels
of N availability served as control plots with no added
amendments. Carbon amendments were used to reduce
soil fertility via stimulation of microbial uptake of
available N, nitrate (NO3), and ammonium (NH4þ)
(Blumenthal et al. 2003). Carbon (as sucrose) was added
at a rate of 252 g Cm2yr1 in three equal 84-g
additions on the same dates as N additions during 2007
and 2008. Soil amendments were sprinkled onto the soil
surface and when possible were added prior to
precipitation events.
Plot design
In October of 2007 and September 2008, C. stoebe
stem and foliar tissues were harvested to provide annual
estimates of biomass. To minimize the effects of stem
clipping on plant growth the following year, removal of
C. stoebe stems and cauline leaves in 2007 occurred after
individual plants had senesced to allow resorption of
nutrients from these tissues. Young rosette leaves that
developed as stems senesced were also left intact. Total
above- and belowground tissues were harvested in
September of 2008. Fine and coarse roots were washed
to remove soil particles, and plant tissues were dried at
608C for five days. Any plants that died before
termination of the experiment were also collected and
processed for biomass and any insect presence.
Experimental blocks were placed 0.75 km apart along
the riparian corridor. To encompass the possible plant
responses to a range of herbivory occurring in the field,
one meadow at the periphery and another at the core of
the infestation were selected for experimental plots. The
lower meadow was near the downstream boundary of
the infestation, with a 10% slope and southeast aspect,
about 40 m uphill from the stream channel. The upper
meadow, at the infestation core, had a 3% slope and
northeast aspect and was 15 m from the stream. Soils
within the upper meadow were coarser in texture and
better drained than those in the lower meadow.
We established 102 plots (17 per treatment) that were
divided equally among the two fenced 40 3 20 m
experimental meadows (51 plots per meadow). Each 0.5
3 0.5 m plot included one randomly selected C. stoebe
rosette and its surrounding vegetation, with most plots
separated by a minimum of 2 m and never situated
downslope from one another. Each plot was a single
replicate in a full-factorial design, with fertilization and
competition manipulations that were spatially randomized within each meadow.
Neighbor removals
Neighboring vegetation was maintained at two levels:
plants were left intact, or 50% of the aboveground
vegetation in each plot (0.125 m2) was clipped at ground
level and removed. Clipping was repeated several times
during the growing season to maintain a reduction in
resident aboveground biomass throughout the experiment. The mean dry mass of aboveground vegetation
removed annually from plots was 509.4 6 79.5 g/m2 in
the upper meadow, and 319.1 6 47.5 g/m2 in the lower
meadow. Values are reported as means 6 SE.
Soil nutrient amendments
Nutrient amendments manipulated soil resource
availability at three levels: nitrogen (N) addition as
ammonium nitrate (high N), no amendments (control),
Plant harvest
Insect sampling
Field densities of Larinus, Urophora, and
Cyphocleonus were not manipulated. Instead, insects
were allowed access to all plots, and we recorded their
occurrence across the imposed resource and competition
gradients of experimental plots. Larinus weevils and
Urophora spp. gall flies found within the seed heads of
target plants were sampled annually. The relative
abundance of these insects at the level of the whole site
was also recorded between 2002 and 2008 (Seastedt et al.
2007, Knochel and Seastedt 2009). For each target plant,
all flower heads were collected and counted in early
August through early October in 2007 and 2008. Seed
heads were sampled as plants senesced to ensure that a
subset of intact capitula was dissected before seed
dispersal. Collections were also timed with the maturation of Urophora and Larinus larvae within flower heads.
Seed production and the number of Urophora and
Larinus larvae, pupae, and adults were recorded by
random dissection of 10 capitula (except in the case of
low flower production) from each target C. stoebe plant.
The number of Cyphocleonus larvae, pupae, adults, or
pupal chambers (indicating past damage) were recorded
by sectioning roots during total C. stoebe biomass
harvest in October 2008.
1906
DAVID G. KNOCHEL AND TIMOTHY R. SEASTEDT
Soil sampling and analyses
To assess the effects of the nutrient manipulations on
soil nutrient availability, soil cores were sampled on a
single date from plots during August 2007, and resin
bags were buried in a subset of plots from May–
September of the 2008 growing season. In late August
2007, three soil cores (2 cm diameter 3 10 cm depth)
were taken and combined in 10 plots (five per meadow)
from each of the three treatment groups (30 replicates).
In neighbor-removal plots, cores were taken from the
clipped portion of the plot, while in control plots,
samples were taken from anywhere within 15–30 cm
from the target C. stoebe plant. Soils were kept cool and
analyzed within 24 hours for extractable N (NO3 þ
NH4þ) using 10 g of soil in a 5:1 2 mol/L potassium
chloride (KCl) to soil ratio extraction (Keeney and
Nelson 1982). To determine gravimetric moisture
content, 10 g of each field-moist soil sample was weighed
wet, then dried at 1058C for three days and reweighed.
In 2008, relative levels of plant-available NO3 and
NH4þ were measured by burying ion-exchange resin
capsules within a subset of representative plots in the
lower meadow (each capsule contains 1 g of ionic
exchange resin beads charged with Hþ and OH ions
held within a porous fabric membrane (UNIBEST PST1, Bozeman, Montana, USA). These resin measurements
served to verify that the soil amendments altered N
availability over the growing season. In May 2008, two
resin capsules per plot were buried 15 cm apart to a
depth of 10 cm, using a 1.5-cm soil coring tube inserted
at a 458 angle. The capsules were buried in four to five
plots within each of the treatment combinations: high or
low N, with neighbors either removed or intact (8–10
bags per treatment). Data from the two capsules per plot
were pooled for analysis, resulting in four to five
replicates per treatment group. Upon collection of resin
capsules in September 2008, soil particles were rinsed
from the surface of each capsule with deionized water,
and resins were refrigerated and analyzed within two
weeks for N concentration (NO3 and NH4þ). Microbial
immobilization during cold storage was likely minimal
(Skogley et al. 1997). Resins were extracted using a
2 mol/L KCl solution, shaken for one hour, and allowed
to sit for 24 hours before extraction. Extracts from soils
or resins were analyzed colorimetrically for inorganic N
with a phenolate assay using a Lachat Automated Ion
Analyzer (Hach, Loveland, Colorado, USA) or an
Alpkem Autoanalyzer (Alpkem RFA Methodology
No. A303-SO21; OI Analytical, College Station, Texas,
USA).
Statistical analyses
SAS PROC MIXED repeated-measures analysis of
covariance (ANCOVA; SAS 2009), was used to analyze
the effects of N availability and neighbor manipulations
or their interaction on spotted knapweed aboveground
biomass, flower, and seed production over the two
growing seasons. The soil N or neighbor-removal
Ecological Applications
Vol. 20, No. 7
treatments were between-subject fixed factors, and year
was the repeated within-subject factor. Additionally, to
control for initial plant size differences, we used the
maximum rosette diameter of plants at the start as a
random covariate in all analyses.
The effects of N or plant competition manipulations
on the attack rates of seed-head insects (per seed head,
per plant, or percentage of seed heads infested) and root
weevil presence (number per root, percentage of roots
infested) were also analyzed with ANCOVA procedures,
and a posteriori Tukey-Kramer tests were used to
compare means. For Larinus and Urophora, the average
number of insects from up to 10 dissected flower heads
per plant was used. Although the flower head insects are
highly mobile, it was unknown whether their oviposition
preferences or attack rates were independent of the
particular plant sampled (insects within plants were
sampled in 2007 before stem harvest, and then again
from new stems produced by the same plants in 2008).
Thus rather than analyze years independently, we took a
more conservative approach and analyzed their cumulative attack rates over two years using repeatedmeasures ANCOVA. The root feeder density, however,
was only sampled and analyzed for its effects on plants
during the 2008 growing season. Unstructured covariance was chosen following Akaike’s Information
Criterion (AIC) and Schwarz’s Bayesian Criterion
(SBC).
Lastly, ANCOVA was used to analyze whether the
impacts of biological control insects (Larinus, Urophora
spp., Cyphocleonus) on seed, flower, or biomass production depended on soil N availability and plant neighbors. For the flower head insects, we used repeatedmeasures ANCOVA to analyze their cumulative effects
over two years. We used PROC GLM (SAS 2009) to
analyze the effects of Cyphocleonus on biomass or
flowering during 2008. The level of root weevil
infestation was standardized and expressed as the
number per unit root biomass. Any response variables
not meeting equality of variance assumptions were logtransformed, and then back-transformed for use in some
tables and figures. For all analyses, statistical significance was determined at P 0.05.
RESULTS
Soil amendments
Analyses of inorganic N levels in soil samples and
buried resin bags confirmed that N amendments
significantly increased N availability. In 2007 soil
samples, fertilized plots had higher and C-amended soils
had significantly lower levels of N (NO3 þ NH4 )
compared with control (ambient) plots, (Appendix:
Fig. A1; F2,20 ¼ 46.13, P , 0.0001). In 2007 soil samples
and 2008 resin bags, neighbor removal did not
significantly affect the N availability of soils beneath
the clipped vegetation compared to control plots. Soil
moisture was also not significantly different in clipped
vs. control plots. However, in the lower meadow during
October 2010
CONTEXT-DEPENDENT CONTROL OF C. STOEBE
1907
FIG. 2. Cumulative effects over two years of (a) nutrient manipulations and (b) plant competition on spotted knapweed
(Centaurea stoebe) flower production (open squares) and aboveground biomass (solid circles) in field plots in Colorado, USA, from
2007 to 2008. Points depict mean 6 SE. Different letters (uppercase for flowers or lowercase for biomass) represent significant
differences between soil nutrient manipulations (across all plots in both meadows) or neighbor-removal treatments (within either
the lower or upper meadow) at P , 0.05. Sample sizes indicate the number of treatment plots, with each plot containing a single C.
stoebe plant.
2007, there was a significant inverse relationship
between soil moisture and soil N availability (R 2 ¼
0.22, F1,29 ¼ 8.05, P ¼ 0.008), while no such relationship
existed in the upper meadow plots. In 2008, resincaptured N was also significantly higher in the fertilized
vs. reduced N treatment (Appendix: Fig. A1; F1,17 ¼
25.84, P , 0.0001). Neighbor removal alone did not
have a significant effect on resin-captured N in field
plots (F1,17 ¼ 0.25, P ¼ 0.62). However, there was a
marginally significant interaction between soil treatment
and neighbor removal (F1,17 ¼ 3.82, P ¼ 0.07), such that
within the fertilized soil treatment, neighbor-removal
plots had lower resin-captured N than control plots, but
no such trend occurred in the low-N treatment.
Soil treatment and neighbor effects on plant growth
Over two growing seasons, flower production and
aboveground biomass of Centaurea stoebe were highest
in the fertilized and in the neighbor-removal plots (Fig.
2). These treatment effects were significant when
comparing the average among both meadows; however,
there was a significant neighbor removal by meadow
interaction such that neighbor removal increased spotted knapweed biomass in the lower meadow (F1,93 ¼
8.88, P ¼ 0.004) but not in the upper meadow (Fig. 2b).
Annual aboveground biomass averaged three times
higher for spotted knapweed plants in the upper
meadow than in the lower meadow (Fig. 2b, Table 1).
Seed number per capitulum did not differ by fertilization
1908
Ecological Applications
Vol. 20, No. 7
DAVID G. KNOCHEL AND TIMOTHY R. SEASTEDT
TABLE 1. Effects of soil and neighbor-removal treatments on Centaurea stoebe (spotted knapweed) response variables (mean 6
SE) in (a) 2007 and (b) 2008 using a general linear model and mixed-model repeated-measures ANCOVA.
Neighbor-removal treatment
Soil N
Measurement
High
a) 2007
Shoot biomass (g)
Flowers per plant
27.2a 6 3.3
66.3a 6 8.5
b) 2008
Total biomass (g)
Lower meadow (g)
Root : shoot ratio
Root biomass (g)
Shoot biomass (g)
Flowers per plant
Seeds per capitulum
31.6a
11.5a
0.66a
6.85a
26.5a
80a
2.31a
6
6
6
6
6
6
6
Lower meadow (boundary)
Control
Low
15.7b 6 2.3 13.0b 6 2.3
41.6ab 6 5.8 34.94b 6 6.3
6.6
14.2b
3.3 7.03ab
0.12 0.62a
1.1
4.45b
6.1
9.7b
21.1 31.3b
0.74 1.84a
6
6
6
6
6
6
6
15.5b
3.10b
0.67a
4.45b
11.6b
31.8b
3.36a
2.3
2.7
0.7
0.7
1.7
6.2
0.6
6
6
6
6
6
6
6
3.9
0.9
0.08
0.9
3.1
9.4
0.9
Upper meadow (core)
Removed
Intact
Removed
Intact
16.3a 6 3.2
47.2a 6 9.1
7.3b 6 1.2
18.9b 6 3.3
21.8 6 2.5
57.6 6 7.6
27.3 6 3.9
63.9 6 9.3
11.3a 6 2.8
3.31b 6 0.8
30.6 6 5.1
33.4 6 7.4
a
0.71
3.50a
7.8a
21.8a
6
6
6
6
b
0.11 1.10
0.6
1.40b
2.2
1.91b
8.0
2.48b
0.27a 6 0.1
6
6
6
6
0.16
0.25
0.61
1.33
0.47
7.96
23.9
70.5
6
6
6
6
0.07
1.1
4.2
12.8
4.46b
0.43 6
7.70 6
27.9 6
89.4 6
6 0.7
0.07
1.3
6.9
23.8
Notes: Different superscript letters denote significant differences within soil or insect treatments detected at P 0.05. Values in
the 2008 lower meadow partial row represent total biomass. There were no significant differences for upper meadow total plant
biomass in 2008 between soil N treatments.
or removal treatments, but was also higher in the
upper meadow than in the lower meadow (Table 1b;
Appendix: Table A1).
Soil treatment and neighbor effects on insect abundance
Over two years, the cumulative abundance of Larinus
minutus per flower head was significantly higher in plots
with neighbors removed compared to control plots
(Table 2; F1,92 ¼ 6.91, P ¼ 0.010). The percentage of
capitula infested was also significantly higher for plants
with neighboring vegetation removed (Table 2; F1,86 ¼
6.87, P ¼ 0.010). Larinus weevil abundances (per seed
head, per plant, or percentage of seed heads infested)
were not significantly affected by soil nutrient amendments. The abundance of Urophora per seed head, per
plant, and percentage of seed heads infested were not
significantly different between any soil or neighborremoval manipulations (Table 2).
For the root weevil Cyphocleonus, 78% of experimental plants (N ¼ 102) were infested, with a range of 0–9
weevils per root, and a mean (6SE) of 2.24 6 0.20; 84%
of roots contained 1–4 weevils. For comparison, roots
sampled from multiple locations along 1 km of the C.
stoebe population in 2008 (N ¼ 60) had an average root
infestation of 82% and a mean (6SE) of 2.00 6 0.27
weevils per root, similar to plants in our experimental
plots. Among the soil N or neighbor-removal treatments, there was not a statistical difference in percentage
of roots infested or in the weevil densities per root
(Table 2). However, infestation by the root weevil was
;20% higher in the high-N plots than in control or lowN plots. Further, Cyphocleonus weevil density per root
had a significantly positive relationship with the plant’s
previous year stem and foliage biomass (R 2 ¼ 0.14,
F1, 101 ¼ 15.64, P , 0.0001).
Herbivory effects across the nutrient
and competition gradients
Prior to harvest, 15.7% (16 plants) of the experimental
plants died. Mortality was significantly greater in the
TABLE 2. Effects of soil N and neighboring plant manipulations on the abundance of L. minutus, Urophora spp., and C. achates
found in the flower heads or roots of Centaurea stoebe plants in field plots in Colorado, USA, 2007–2008.
Urophora spp. Treatment
High N
Control
Low N
Neighbors removed
Neighbors intact
Gall flies per
capitulum
0.84
0.75
0.91
0.77
0.89
6
6
6
6
6
0.05
0.05
0.07
0.04
0.05
Capitula
infested
(%)
27
34
27
30
29
6
6
6
6
6
4
4
4
4
3
Larinus minutus Gall flies
per plant§
Weevils per
capitulum
Capitula
infested
(%)
17
11
9
8
17
0.64
0.59
0.51
0.65a
0.50b
56
52
45
57a
45b
6
6
6
6
6
3.5
2.2
1.7
1
2.8
6
6
6
6
6
0.04
0.04
0.04
0.03
0.03
6
6
6
6
6
4
4
5
3
4
Cyphocleonus achatesà
Weevils
per plant
Weevils per
plant root
43a
18b
16b
32a
18b
2.76
2.14
1.85
2.45
2.04
6
6
6
6
6
5.2
2.1
2.9
3.3
2.8
6
6
6
6
6
0.33
0.37
0.32
0.27
0.29
Plants
infested
(%)
91
72
73
71
86
Notes: Different superscript letters denote significant differences within soil N or removal treatments detected at P , 0.05.
Values are means 6 SE. The insects sampled included a foliage and seed feeder (the weevil L. minutus), gall flies (Urophora spp.),
and a root feeder (the weevil Cyphocleonus achates).
Values were calculated using the average number from a subset of dissected capitula for each experimental plant.
à Roots were sampled only in 2008.
§ Estimated as the product of the mean number per capitulum, percentage of capitula infested, and number of capitula per plant.
October 2010
CONTEXT-DEPENDENT CONTROL OF C. STOEBE
lower meadow (13 plants) than in the upper meadow (3
plants) (v2: df ¼ 3, N ¼ 105; P ¼ 0.012), and 50% of the
plants that died contained root weevils (between one and
three larvae per root). The numbers of dead plants in the
high-N, control, and low-N plots were 3, 6, and 7,
respectively; while mortality in plots with neighbors
removed vs. neighbors intact were 6 and 10, respectively.
Mortality between soil N and neighbor-removal treatments was not significantly different than expected if
mortality were random.
Root biomass and Cyphocleonus density were highly
positively correlated (i.e., larger roots contained more
weevils; F1, 101 ¼ 20.30, P , 0.0001). Thus weevil
numbers were standardized per unit root biomass to
analyze the impact of root herbivory on the plant
response. Further, roots containing no weevils (N ¼ 22)
were excluded from the analysis in order to isolate plant
responses when weevils were present. Soil nutrient and
neighbor treatments, as well as initial plant size, did not
explain additional variance in biomass or flower
production, so these parameters were removed. In the
resulting regression model, Cyphocleonus density was a
significant predictor of 2008 flower production (F1,76 ¼
60.96, P , 0.0001) and aboveground biomass (F1,76 ¼
67.80, P , 0.0001). Flower production and aboveground biomass were inversely related to Cyphocleonus
root density (Fig. 3a, b).
Seed production in both years was negatively correlated with the presence of the seed-head weevil, Larinus
(Table 3), such that on average, for every additional
weevil within a capitulum, plants produced 5.1 6 1.7
(mean 6 SE) fewer seeds. Average seed production and
insect presence also varied by year (Table 3). The
negative correlation between seeds and Larinus was
stronger in plots with intact vegetation and marginally
significant in plots with neighbors removed (intact:
Pearson’s r ¼ 0.41, F1,26 ¼ 4.83, P ¼ 0.038; removal:
Pearson’s r ¼0.32, F1,32 ¼ 3.33, P ¼ 0.078). Considering
soil N manipulations, the negative slope describing the
relationship between seeds per capitulum and Larinus
presence was greatest in low-N plots (Pearson’s r ¼
0.52, F1,15 ¼ 5.28, P ¼ 0.038), marginally significant in
fertilized plots (Pearson’s r ¼ 0.44, F1,19 ¼ 4.02, P ¼
0.061), and not significant in control plots (F1,23 ¼ 0.98,
P ¼ 0.333).
In 2007, there was a positive correlation between seed
production and Urophora spp. presence (Table 3), such
that for every additional fly within a capitulum, plants
produced 8.5 6 1.7 (mean 6 SE) more seeds. However,
when comparing 2007 seed production to the combined
presence of Larinus and Urophora across all dissected
seed heads, seeds showed only a slight positive
correlation with Urophora (Pearson’s r ¼ 0.07), in a
model controlling for the presence of a strong negative
relationship with Larinus (Pearson’s r ¼ 0.30). A
similar model combining both insect effects for 2008
found that Urophora was not significantly correlated
with seed production above and beyond the negative
1909
FIG. 3. Negative logarithmic regressions of (a) flower
production (originally measured as flowers/plant) and (b)
aboveground biomass (originally measured in grams) as a
function of the density of Cyphocleonus achates root weevils per
unit root biomass in the 2008 study year.
effects of Larinus (Pearson’s r, Larinus ¼ 0.36; overall
model, R 2 ¼ 0.13, F2, 474 ¼ 36.80, P , 0.001). Within
individual experimental plants, Urophora and Larinus
presence were negatively correlated (Pearson’s r ¼0.22,
F1, 103 ¼ 5.63, P , 0.0196), and their relationship was of
similar strength and direction when compared across all
dissected capitula in 2007 and 2008.
DISCUSSION
The results demonstrate that a combination of
biological control insects reduce biomass and seed
production in Centaurea stoebe, and conditions that
lower resource availability, such as the presence of plant
competition, can intensify the effects of herbivory and
decrease plant performance. Consistent with predictions
of the herbivory–resource gradient model (Fig. 1),
resource availability and plant competition governed
the response of C. stoebe to herbivory from biological
control insects. However, the importance of the resource
or competition environments in moderating these
responses varied by insect species. For Cyphocleonus,
as the intensity of damage from weevils to the taproot
increased, aboveground biomass and flower production
declined, irrespective of resource or competition condi-
1910
Ecological Applications
Vol. 20, No. 7
DAVID G. KNOCHEL AND TIMOTHY R. SEASTEDT
TABLE 3. Abundance and relationship between seeds, Larinus, and Urophora found in flower heads of Centaurea stoebe in field
experimental plots.
No. seeds/flower head
Larinus minutus
Year and
total no. seeds
Overall
Larinus
absent
Larinus
present
2007 (N ¼ 905 seeds)
2008 (N ¼ 517 seeds)
10.2 6 0.3
4.1 6 0.3
13.9 6 0.5
7.7 6 0.6
6.1 6 0.3
1.8 6 0.2
Correlation
with seeds
r
0.35***
0.31*
R2
No. per
flower head
Capitula
infested (%)
0.12
0.09
0.54 6 0.02 (0–3)
0.65 6 0.03 (0–4)
48
60
Notes: Values are means 6 SE (with range in parentheses for insect numbers) calculated from a subset of dissected capitula per
experimental plant. Correlations are Pearson’s r values in regression analyses. Significance is indicated by asterisks. R 2 values are
from regression analyses testing the influence of L. minutus or Urophora spp. on seed production.
*P 0.05; ***P , 0.001; ns, not significant (P . 0.05).
tions (Fig. 3). Thus under field conditions, higher root
weevil densities exerted substantial control on spotted
knapweed fitness and appeared to supersede any benefits
of a favorable soil resource environment or reduced
plant competition.
In contrast, while the Larinus weevil reduced seed
production over all treatments and during both years
(Table 3), there was some indication that the weevil’s
negative effects on the plant were strengthened when soil
resources were reduced or when plant competition was
increased, at least under the range of conditions tested
here. This fits the prediction that spotted knapweed may
produce more seed when resources are abundant, but
nonetheless remains constrained by Larinus, especially
when resources are less readily available. The direct
influence of Urophora gall flies on seed production was
negligible, and we found no evidence that the external
environmental conditions moderated their influence on
the plant. The positive correlation in 2007 between the
numbers of seed and Urophora per seed head is
explained through interactions with Larinus weevils.
Larinus consumes seeds and Urophora developing within
a flower head (Seastedt et al. 2007). As a result, when
Larinus is absent from a flower head, both seeds and
Urophora may increase in number. Thus the overall
observed plant responses to herbivory varied by insect
species, and in the case of Larinus, fit the prediction that
resources and plant competition moderate the response.
Moreover, as we discuss, resource limitations and the
presence of plant competition exerted direct negative
influences on spotted knapweed growth, and in essence
reduced the baseline point from which plants begin the
predicted compensation responses to damage from these
biological control insects.
In the hypothesized herbivory–resource gradient
model (Fig. 1), it was first necessary to test the
assumption that spotted knapweed is negatively influenced by reduced resource availability and plant
competition in field conditions. Indeed, C. stoebe shoot
biomass more than doubled in fertilized plots (Table 1),
verifying that higher N conditions favor adult spotted
knapweed growth, similar to the responses demonstrated by Story et al. (1989). Soil N and C amendments had
significant effects on inorganic N, but fertilized plots
differed in N content in much greater magnitude from
control plots than did the reduced-N plots. Nearly all
biomass and growth measurements differed when
comparing the fertilized treatment to control and lowN plots, but plant measurements in low-N plots did not
differ from the control (Table 1). C. stoebe is thus able to
increase growth substantially in response to high
nutrient availability, but is meanwhile also tolerant of
lower N conditions. These results support the observation that while spotted knapweed generally grows best
and dominates highly disturbed sites such as riparian
channels, roadsides, or areas high in resources, the weed
can also maintain populations within intact rangelands
that may be lower in nutrient availability.
Plant competition exerted large restraints on C. stoebe
growth only in the lower meadow, as evidenced by a
greater than threefold increase in root and shoot
biomass and flower production in plots where neighboring plant cover was reduced (Fig. 2b). We are
uncertain why neighbor removal did not affect growth in
the upper experimental meadow. We observed dissimilarities among the resident plant communities, soil
properties, and Larinus attack between the two areas,
and these factors may have contributed to the contrasting responses to neighbor removal, although we lacked
sufficient data to adequately test their effects on target
plants. For example, in the lower meadow C. stoebe
plants were smaller, and the resident native and
nonnative plant community (the removed neighbors)
comprised mostly graminoids, while in the upper
meadow C. stoebe produced much greater biomass and
made up a higher percentage of the vegetation removed
beside study individuals. Soil N availability measured
within all plots was not different between the two
meadows, although we found a negative relationship
between soil moisture and soil N availability in soils
sampled from the lower meadow in 2007, suggesting
interaction between the two. Finally, attack by Larinus
weevils (number per capitulum, percentage of capitula
infested) was significantly higher in the lower meadow
than in the upper meadow; however, we did not quantify
adult weevil damage to foliage. Recent work by Maron
and Marler (2008) showed that invading seedlings of C.
stoebe were unaffected by established native competi-
October 2010
CONTEXT-DEPENDENT CONTROL OF C. STOEBE
CONCLUSION
TABLE 3. Extended.
Urophora spp.
Correlation
with seeds
r
0.37***
0.15 ns
1911
R2
No. per
flower head
Capitula
infested (%)
0.14
0.02
0.27 6 0.02 (0–4)
0.95 6 0.06 (0–5)
22
49
tion, yet here we show that native competition is capable
of having large effects on spotted knapweed once
established, particularly when multiple species of biological control insects are present. In summary, while
resource limitation and competition were not found to
consistently affect how spotted knapweed responded to
herbivory, these factors nonetheless restrained plant
growth and reproduction, and represent conditions that
amplify the effects of biological control insects.
In addition to observations of extensive damage and
increased mortality due to root, stem, and foliage
herbivory, the monitoring effort at this field site has
found decreasing seed production over the 2003–2008
interval (Knochel and Seastedt 2009). Considering the
results described here, these reductions in reproductive
output over several years are strongly governed by
interactions of insect, nutrient, and plant competition
effects (Knochel 2009). This evidence, in addition to
multiple studies reporting on the capacity of species and
functional group diversity to resist invasion by spotted
knapweed at realistic propagule densities (e.g., Pokorny
et al. 2005), are consistent with the hypothesis that
herbivory and plant competition for limiting resources
are capable of greatly reducing the threats of this
invasive species.
Summarizing the questions posed here, we first
showed that reduced levels of nitrogen availability, and
competition from neighboring vegetation, can greatly
reduce the growth of this dominant invasive plant.
Secondly, we found that the seed-head weevil, Larinus,
was more abundant on knapweed plants where adjacent
vegetation was removed and where spotted knapweed
densities were also lower. This indicates that Larinus
may congregate on more apparent plants or on residual
patches on the periphery of infestations, thereby
potentially decreasing plant spread through greater
damage and depletion of seed production on the edges
of infestations. Meanwhile, resource availability and
plant competition do not affect attack rates of the rootfeeding weevil, Cyphocleonus. We confirmed that the
negative effects of the root-feeding weevil on plant
growth and reproduction observed in an experimental
garden study (Knochel 2009) also occurred in the field at
realistic insect and plant densities, and that this root
weevil can have negative effects on plant fitness that
supersede any benefits gained when growing in a more
plentiful resource environment.
Earlier studies that identified benefits of herbivores to
plant fitness (Callaway et al. 1999, Thelen et al. 2005,
Newingham et al. 2007, Pearson and Callaway 2008) did
so using single herbivores, and using experimental
conditions that likely affected the resource environments
in which the studies were conducted. An opposite
conclusion is presented here, and our findings join other
studies reporting negative impacts (reviewed in Knochel
and Seastedt 2009). Clearly, herbivore impacts vary
across a range of plant competition and N availability
(and undoubtedly to other variables not studied here).
However, under a number of limited-resource conditions and using multiple species of biological control
agents, we predict moderate to strong negative effects of
herbivory on C. stoebe and therefore expect these species
to reduce densities and impacts of C. stoebe in many
areas of North America.
ACKNOWLEDGMENTS
We thank Christine Fairbanks, Kyle Wald, Kali Blevins,
Justin Feis, Irene Hale, and Nataly Ascarrunz for field and lab
help. We also thank Linda and Sergio Sanabria for use of their
land to conduct research. Deane Bowers, William Bowman,
Carol Wessman, and Susan Beatty provided useful suggestions
and comments for improving the manuscript. This work was
funded by the National Research Initiative of the USDA
Cooperative State Research, Education and Extension Service,
grant number 06-03618.
LITERATURE CITED
Blumenthal, D. M., N. R. Jordan, and M. P. Russelle. 2003.
Soil carbon addition controls weeds and facilitates prairie
restoration. Ecological Applications 13:605–615.
Callaway, R. M., T. H. DeLuca, and W. M. Belliveau. 1999.
Biological-control herbivores may increase competitive ability of the noxious weed Centaurea maculosa. Ecology 80:
1196–1201.
Chobot, V., C. Huber, G. Trettenhahn, and F. Hadacek. 2009.
(6)-Catechin: chemical weapon, antioxidant, or stress
regulator? Journal of Chemical Ecology 35:980–996.
Corn, J. G., J. M. Story, and L. J. White. 2006. Impacts of the
biological control agent Cyphocleonus achates on spotted
knapweed, Centaurea maculosa, in experimental plots.
Biological Control 37:75–81.
Corn, J. G., J. M. Story, and L. J. White. 2007. Effect of
summer drought relief on the impact of the root weevil
Cyphocleonus achates on spotted knapweed. Environmental
Entomology 36:858–863.
Cortilet, A. B., and N. Northrop. 2006. Biological control of
European buckthorn and spotted knapweed. Minnesota
Department of Agriculture, Final Program Report,
St. Paul, Minnesota, USA. hhttp://www.mda.state.
mn.us/news/ publications/pestsplant s/weedc ontrol/
knapweedlcmrfinalreport.pdfi
Duke, S. O., A. C. Blair, F. E. Dayan, R. D. Johnson, K. M.
Meepagala, D. Cook, and J. Bajsa. 2009. Is ()-Catechin a
novel weapon of spotted knapweed (Centaurea stoebe)?
Journal of Chemical Ecology 35:141–153.
Hufbauer, R. A., and R. Sforza. 2008. Multiple introductions
of two invasive Centaurea taxa inferred from cpDNA
haplotypes. Diversity and Distributions 14:252–261.
Jacobs, J. S., S. E. Sing, and J. M. Martin. 2006. Influence of
herbivory and competition on invasive weed fitness: observed
effects of Cyphocleonus achates (Coleoptera: Curculionidae)
1912
DAVID G. KNOCHEL AND TIMOTHY R. SEASTEDT
and grass-seeding treatments on spotted knapweed performance. Environmental Entomology 35:1590–1596.
Keeney, D. R., and D. W. Nelson. 1982. Nitrogen: inorganic
forms. Pages 643–698 in A. L. Page, R. H. Miller, and D. R.
Keeney, editors. Methods of soil analysis. Agronomy
Monograph 9, Part 2. Second edition. American Society of
Agronomy, Madison, Wisconsin, USA.
Knochel, D. G. 2009. Interactions among herbivory, soil
resources, and plant competition govern the recruitment
and dominance of spotted knapweed (Centaurea stoebe) in
North America. Dissertation. University of Colorado,
Boulder, Colorado, USA.
Knochel, D. G., and T. R. Seastedt. 2009. Sustainable control
of spotted knapweed, (Centaurea stoebe). Pages 211–225 in
Inderjit, editor. Management of invasive weeds. Springer
Science þ Business Media B. V., New York, New York, USA.
Maron, J. L., and M. Marler. 2008. Field-based competitive
impacts between invaders and natives at varying resource
supply. Journal of Ecology 96:1187–1197.
Maschinski, J., and T. G. Whitham. 1989. The continuum of
plant responses to herbivory: the influence of plant association, nutrient availability, and timing. American Naturalist
134:1–19.
McNaughton, S. J. 1979. Grazing as an optimization process:
grass ungulate relationships in the Serengeti. American
Naturalist 113:691–703.
Michels, G. J., Jr, V. A. Carney, D. Jurovich, S. Kassymzhanova-Mirik, E. Jones, K. Barfot, J. Bible, M. Mirik, S. Best,
E. Bustos, B. Karl, and D. Jimenez. 2009. Biological control
of noxious weeds on federal installations in Colorado and
Wyoming. Texas Agricultural Experiment Station
2009 Annual Report. hhttp://amarillo.tamu.edu/programs/
agrilife_programs/entomology_research/entomology_research_
colorado_weed_biocontrol.phpi
National Resources Conservation Service. 2008. Web Soil
Survey. U.S. Department of Agriculture, Washington, D.C.,
USA. hhttp://websoilsurvey.nrcs.usda.gov/i
Newingham, B. A., R. M. Callaway, and H. BassiriRad. 2007.
Allocating nitrogen away from an herbivore: a novel
compensatory response to root herbivory. Oecologia 153:
913–920.
Ochsmann, J. 2001. On the taxonomy of spotted knapweed.
Pages 33–41 in L. Smith, editor. Proceedings of the First
International Knapweed Symposium of the Twenty-first
Century (Coeur d’Alene, Idaho, USA, 2001). USDA-ARS,
Albany, California, USA.
Pearson, D. E., and R. M. Callaway. 2008. Weed biocontrol
insects reduce native plant recruitment through second-order
apparent competition. Ecological Applications 18:1489–
1500.
Pearson, D. E., and R. J. Fletcher. 2008. Mitigating exotic
impacts: restoring deer mouse populations elevated by an
exotic food subsidy. Ecological Applications 18:321–334.
Ecological Applications
Vol. 20, No. 7
Pokorny, M. L., R. L. Sheley, C. A. Zabinski, R. E. Engel, T. J.
Svejcar, and J. J. Borkowski. 2005. Plant functional group
diversity as a mechanism for invasion resistance. Restoration
Ecology 13:448–459.
SAS Institute. 2009. Statistical analysis system, proprietary
software release. Version 9.2. SAS Institute, Cary, North
Carolina, USA.
Seastedt, T. R., D. G. Knochel, M. Garmoe, and S. A. Shosky.
2007. Interactions and effects of multiple biological control
insects on diffuse and spotted knapweed in the Front Range
of Colorado. Biological Control 42:345–354.
Shea, K. D., D. Kelly, A. W. Sheppard, and T. L. Woodburn.
2005. Context-dependent biological control of an invasive
thistle. Ecology 86:3174–3181.
Sheley, R. L., J. S. Jacobs, and M. L. Carpinelli. 1999. Spotted
knapweed. Pages 350–361 in R. L. Sheley and J. K. Petroff,
editors. Biology and management of noxious rangeland
weeds. Oregon State University, Corvallis, Oregon, USA.
Skogley, E. O., A. Dobermann, J. E. Yang, B. E. Schaff,
M. A. A. Adviento, and M. F. Pampolino. 1997. Methodologies for resin capsules: capsule storage and ion recovery.
Sciences of Soils. Release 2. hhttp://hintze-online.com/sos/
1997/Articles/Art3/i
Story, J. M., K. W. Boggs, and E. R. Graham. 1989. Effects of
nitrogen fertilization on spotted knapweed and competing
vegetation in western Montana. Journal of Range Management 42:222–225.
Story, J. M., W. Callan, J. G. Corn, and L. J. White. 2006.
Decline of spotted knapweed density at two sites in western
Montana with large populations of the introduced root
weevil, Cyphocleonus achates (Fahraeus). Biological Control
38:227–232.
Story, J. M., and G. L. Piper. 2001. Status of biological control
efforts against spotted and diffuse knapweed. Pages 11–17 in
L. Smith, editor. Proceedings of the First International
Knapweed Symposium of the Twenty-first Century (Coeur
d’Alene, Idaho, USA, 2001). USDA-ARS, Albany, California, USA.
Story, J. M., L. Smith, J. G. Corn, and L. J. White. 2008.
Influence of seed head–attacking biological control agents on
spotted knapweed reproductive potential in western Montana over a 30-year period. Environmental Entomology 37:
510–519.
Thelen, G. C., J. M. Vivanco, B. A. Newingham, W. Good,
H. P. Bais, P. Landres, A. J. Caesar, and R. M. Callaway.
2005. Insect herbivory stimulates allelopathic exudation by
an invasive plant and the suppression of natives. Ecology
Letters 8:209–217.
Wise, M. J., and W. G. Abrahamson. 2007. Effects of resource
availability on tolerance of herbivory: a review and
assessment of three opposing models. American Naturalist
169:443–454.
APPENDIX
Inorganic N levels from field plots in 2007 and 2008, and ANOVA results of Centaurea stoebe measurements during 2007, 2008,
and the entire experimental period (Ecological Archives A020-070-A1).