Additive effects of aboveground and belowground herbivores on the dominance of spotted knapweed (Centaurea stoebe)

Transcription

1 DOI /s z PLANT-ANIMAL INTERACTIONS - ORIGINAL PAPER Additive effects of aboveground and belowground herbivores on the dominance of spotted knapweed (Centaurea stoebe) David G. Knochel Nathan D. Monson Timothy R. Seastedt Received: 11 December 2009 / Accepted: 17 June 2010 Ó Springer-Verlag 2010 Abstract Spotted knapweed (Centaurea stoebe) is found in over 3 million ha of rangeland and forests across North America, and evidence supporting the use of biological control as a regional method to reduce infestations and their associated impacts remains inconclusive. Several species of insects have been reported to reduce plant densities in some areas; however, rigorous studies that test combinations of these species and the influence of resource availability are lacking. We examined the singular and combined effects of herbivory by a root weevil (Cyphocleonus achates) and a flower head weevil (Larinus minutus) on the growth and flower production of C. stoebe. We also manipulated soil resource fertility as an additional factor that could explain the outcomes of contradictory biological control herbivore effects on C. stoebe. In a greenhouse study, herbivory by C. achates decreased flower production for plants across all resource environments. In a caged common garden study, the negative effects of herbivory also did not interact with soil nutrient status. However, the presence of plant competition further Communicated by Debra Peters. Electronic supplementary material The online version of this article (doi: /s z) contains supplementary material, which is available to authorized users. D. G. Knochel (&) T. R. Seastedt Department of Ecology and Evolutionary Biology, Institute of Arctic and Alpine Research, University of Colorado at Boulder, Campus Box 450, Boulder, CO , USA David.Knochel@colorado.edu N. D. Monson Department of Ecology, Evolution, and Marine Biology, University of California, Santa Barbara, USA decreased knapweed growth, and the negative effects of concurrent herbivory by C. achates and L. minutus on plant biomass and flower production were additive. Derived within the context of variable levels of soil nutrient availability and competing vegetation, these results support the cumulative stress hypothesis and the contention that combined above- and belowground herbivory can reduce spotted knapweed densities and reduce the ecological and economic impacts of this species in rangelands of western North America. Keywords Biological control Herbivory Compensation Plant competition Nutrient limitation Introduction A better understanding of above- and belowground biotic and abiotic processes that influence the dominance of nonnative invasive plant species could offer additional insights into mechanisms of their sustainable control. A longstanding view is that herbivores have greater negative impacts on plant fitness in conditions of lower soil resources (Mattson 1980). A recent model by Wise and Abrahamson (2007) predicts that resource limitation in plants for defense, tolerance, or to compensate for herbivore damage regulates plant herbivore interactions. The model predicts that if an herbivore directly reduces a plant s ability to capture the focal resource limiting its growth (e.g., N, light, etc.), then the ultimate outcome of a plant herbivore interaction will be detrimental to the plant. In the context of spotted knapweed [Centaurea stoebe L. subsp. micranthos (Gugler) Hayek, also identified as C. maculosa and C. biebersteinii; see Ochsmann 2001; Hufbauer and Sforza 2008), the limiting resource model predicts that fitness outcomes of herbivore

2 consumption of foliar and root tissues may depend on whether the loss of those tissues ultimately reduces the plant s ability to obtain a limiting resource. Spotted knapweed occupies over 3 million ha of land in North America, and is probably the most widely distributed and problematic of the Eurasian knapweeds (Centaurea spp.) that have become dominant components of the rangeland vegetation in North America (Sheley et al. 1999). The plant has been linked to a wide range of ecological and economic problems that include soil erosion and loss of forage (e.g., Lacey 1989; Duncan et al. 2004). A combination of cultural, mechanical, and chemical control methods have been employed over the past several decades in an attempt to control this species (Sheley et al. 1999). In addition, a collection of 13 species of biological control insects were released in efforts to reduce the abundance of C. stoebe and C. diffusa (Story and Piper 2001). These insects have controlled diffuse knapweed across much of its introduced range (Seastedt et al. 2003, 2007; Coombs et al. 2004; Smith 2004; Myers et al. 2009). However, reports of these species successfully reducing spotted knapweed populations (Corn et al. 2006; Cortilet and Northrop 2006; Jacobs et al. 2006; Story et al. 2006, 2008; Michels et al. 2009) are debated (e.g., Pearson and Ortega 2009), and the release of some insect species has led to unintended negative consequences (Pearson and Fletcher 2008). Understanding the influence of simultaneous above- and belowground damage (Masters et al. 1993; Maron 1998; van Ruijven et al. 2005; van der Putten et al. 2009) is an essential step in determining whether a combination of herbivores can sustainably reduce populations of an invasive plant. For example, spotted knapweed s stem and cauline leaves provide additional surface area for photosynthesis and carbon fixation once basal rosette leaves become shaded by adjacent vegetation or senesce (Hill and Germino 2005). If the flower head weevil, Larinus minutus Gyll. (Coleoptera: Curculionidae), consumes these critical tissues during bolting, the plant may be more severely limited by an inability to produce adequate photosynthates. These effects would be in addition to the well-studied and large negative effects of L. minutus larvae on seed production (Story et al. 2008; Knochel and Seastedt 2009, 2010; Knochel et al. 2010). Likewise, Cyphocleonus achates Fahr. (Coleoptera: Curculionidae) root weevil larvae can heavily damage taproot tissues in spring to early summer. In theory, this damage could reduce nutrient capture or diminish the advantages afforded by the plant s taproot to access deeper soil water during a critical period of growth (Hill et al. 2006). Damage by root feeding and emergence could also increase plant susceptibility to microbial pathogens. In combination, damage by these two herbivores might further reduce fitness more than either insect alone, and their influence could also potentially interact with background resource availability. Evidence suggests that spotted knapweed can actually benefit (e.g., produce more tissue in the presence of lowlevel damage compared to undamaged plants) from root herbivory or low-intensity clipping under some conditions (Kennett et al. 1992; Ridenour and Callaway 2003; Thelen et al. 2005; Newingham et al. 2007). In contrast, studies with more intense defoliation by multiple species, or herbivore damage concurrent with plant competition or stressful environmental conditions, have produced neutral to negative compensation responses to herbivory (e.g., Müller-Schärer 1991; Kennett et al. 1992; Steinger and Müller-Schärer 1992; Jacobs et al. 2006). To further complicate the issue, field studies indicate that a persistent drought in the western US may have been principally responsible for some spotted knapweed declines (Sturdevant et al. 2006; Pearson and Fletcher 2008; Pearson and Ortega 2009), or may have interacted with insect effects (Corn et al. 2007). Taken together, the degree to which a combination of introduced insects, plant competition, and background environmental conditions will influence spotted knapweed densities across the majority of invaded rangelands and pastures in North America remains unclear. The limiting resource model (Wise and Abrahamson 2007) thus provides a suitable hypothetical context under which to study these plant herbivore interactions in multiple resource conditions. The goal of the present study was to improve understanding of the singular and combined effects of two biological control insects on spotted knapweed, C. achates and L. minutus, and elucidate whether their influence varies across a gradient in soil nutrient availability. Specifically, we measured the extent to which C. achates and L. minutus could impact spotted knapweed growth and flower production, and we asked three questions relevant to the efficacy of these insects: 1. Can spotted knapweed compensate or overcompensate for realistic levels of root or aboveground damage from these herbivores? 2. How does soil nutrient availability affect spotted knapweed herbivore interactions? 3. Do C. achates and L. minutus in combination have greater effects than either species alone? Methods Greenhouse experiment Research was conducted at the University of Colorado at Boulder greenhouse. In May 2007 we filled sixty 7cm9 40 cm tree seedling containers with a 1:1 mixture of FAFARD Growing Mix #2 (70% peat with perlite and

3 vermiculite) and loamy clay field soil, and packed the containers into racks that were buried up to the soil surface outdoors and rotated monthly. Four C. stoebe seeds were planted per pot, thinned to one per container, and grown for four months. In early October 2008, we collected adult C. achates root weevils from field populations near Boulder, CO, USA, and three male and eight female adults were placed among the C. stoebe rosettes and enclosed with a metal corral using 30 cm tall flashing that was buried 5 cm deep. The flightless C. achates weevil feeds on knapweed stem and foliage tissues during the growing season, females oviposit at the root crown, and larvae mine into the tap root and typically enter diapause during first instar (Corn et al. 2009). In the spring, larvae continue feeding within the root, and at fourth instar form a pupal chamber with adults emerging mid-summer. Although C. achates may select larger plants for oviposition under field conditions (Knochel and Seastedt 2010), we standardized initial plant size (rosettes were cm) and thus expected the resulting larval densities to be random and oviposition to occur across the majority of the plants. Ten rosettes were grown outside of the metal corral to ensure that some plants would remain uninfested. The plants were transferred to 3.8 L pots on 6 November and moved inside the greenhouse, where the conditions allowed the larvae to forego winter diapause. All plants were watered regularly and rotated on a weekly basis to ensure homogeneous greenhouse light and temperature conditions across treatments. We also used supplemental light to induce flowering and to approximate the 14 h of daylight during May in Colorado. Plants were randomly assigned to one of two soil treatments, with approximately 28 plant replicates per treatment: ambient soil resources (control), or a reduced level of inorganic nitrogen (N) availability (low N). Carbon additions were used to reduce soil fertility by stimulating microbial uptake of soluble inorganic N (Blumenthal et al. 2003). Half of the pots (N = 28) received three sucrose additions equivalent to a total of 256 g C m -1 year -2 (4 g on 17 November 2007, 6 g on 14 December 2007, and 8 g on 4 February 2008). The sucrose was sprinkled onto the soil surface and lightly watered. To confirm that C amendments decreased available N, we collected four soil samples of mixed, homogenized soil from the control and low-n treatments at the end of the experiment on 21 February Soils were processed within 24 h for N content (NO 3 -? NH 4? ) using 10 g of soil in an extraction employing a 5:1 ratio of 2 M KCl to soil. Extracts were analyzed colorimetrically for inorganic N with a phenolate assay using an Alpkem autoanalyzer (Alpkem Corp. RFA Methodology No. A303-SO21) (Keeney and Nelson 1982). Plants initiated stem growth on 5 December 2007, and began flowering in mid-january. On 28 December, four plants were harvested from the low-n treatment group that died while bolting. No attempts to pollinate the flowers were made and therefore seed production was not recorded. We report field data on seed production in relation to C. achates density in Knochel and Seastedt (2010). The first adult C. achates emerged from its host plant on 04 February 2008, approximately three months after oviposition. Plants were harvested on 21 February Numbers of capitula per plant were recorded, and taproots were examined for the quantity of C. achates larvae, pupae, or adults. Fine and coarse roots were washed to remove soil particles. All above- and belowground plant tissues were dried at 60 C for five days and weighed. To determine tissue N content, the root and shoot samples were pulverized with a Wiley mill (0.25 mm 2 openings, 40 mesh screen), and 3 5 mg per plant were analyzed by combustion on a Carlo Erba Flash EA1110 CN analyzer (CE Elantech, Lakewood, NJ, USA). Statistical analysis Spotted knapweed root infestation numbers and soil treatment effects on available N and tissue N were analyzed using analysis of variance (ANOVA) procedures, and means were compared with a posteriori Ryan Einot Gabriel Welsch multiple-range tests (SAS v. 9.2, 2009). ANCOVA procedures were used to determine the effects of C. achates density on plant traits under the different resource conditions, with N level used as a fixed factor and number of C. achates per root used as the continuous covariate. In an analysis of flower production, shoot biomass was included as an additional covariate to test for effects of C. achates while holding plant size constant. Variables were log-transformed if necessary to fulfill assumptions of normality of variance, and then backtransformed for clarity and the relevance of values within figures. Common garden experiment The experiment was conducted for two years between August 2006 and September 2008 on a grassland at the University of Colorado East Campus Research Park, in Boulder, CO, USA ( N, W, elevation 1,604 m). Mean annual air temperature was 9 11 C, with a frost-free period of days (NOAA 2009). The soils on the site are clay-loams and are moderately compacted. The site has a slight slope (2%) with an easterly aspect and full sun (NRCS 2009). During the course of this experiment, the site received 94.5% of its average precipitation (30 year average precipitation recorded in

4 Boulder, CO, USA; NOAA 2009), of which 57% normally falls between April and August. Despite the near-average yearly precipitation during the experiment, the 2007 and 2008 growing season precipitations (April September) were 66.8 and 88.3% of the average, indicating drier than average summers and wetter than average winters. The existing plant community included the native grasses Pascopyrum smithii (Rydb.) A. Löve (Poaceae), Bouteloua gracilis (Willd. ex Kunth) Lag. ex Griffiths (Poaceae), and the native forb Erigeron flagellaris A. Gray (Asteraceae). Non-native species represented about half of the extant plant cover, and common species included the grasses Agropyron cristatum (L.) Gaertn. (Poaceae) and Poa pratensis L. (Poaceae), and the non-native dicots Convolvulus arvensis L. (Convolvulaceae), Lactuca serriola L. (Asteraceae), Alyssum minus (L.) Rothm (Brassicaceae), Brassica rapa L. (Brassicaceae), Plantago lanceolata L. (Plantaginaceae), and Erodium cicutarium (L.) L Hér. ex Aiton (Geraniaceae). To equalize the background plant cover, the area was initially mowed to 8 cm and clippings were removed. One hundred forty-four 0.25 m 2 plots were then delineated in a m area, and separated by a 0.5 m buffer zone. Each plot contained a caged C. stoebe rosette that was transplanted into the existing plant community, and represented a single randomized replicate of a full-factorial design with fertilization (three levels) and insect (four levels) manipulations, and with 12 replicates per treatment. Soil resource availability was held at three levels: nitrogen (N) addition as ammonium nitrate (high N), no amendments (control), and carbon (C) addition (as sucrose) to reduce N availability (low N). Biological control insects were manipulated within cages at four levels: (1) both the root weevil C. achates and the flower head weevil L. minutus; (2) C. achates only; (3) L. minutus only, or; (4) no insects. Experimental C. stoebe target plants were grown from seed collected in 2002 from a single field population located in the foothills 10 km north of Boulder, CO, USA. On 1 May 2006 we prepared plants according to the procedures used in the greenhouse experiment, and after seven weeks each soil core containing a mature rosette and its central taproot were transplanted into the plots. At transplant, the maximum rosette diameter was measured for use as a covariate in the statistical analyses. To ensure initial survival, transplants were watered from July to September of 2006, and any mortality during this period was ascribed to transplanting stress and the plant was replaced. Plants also received supplemental water during the months of April July 2007, when precipitation dropped below 80% of the 30-year average. Thus, plant responses to insect and nutrient treatments were monitored under average moisture conditions for the Colorado Front Range. Each C. stoebe plant and the adjacent vegetation within a 0.25 m radius were enclosed with a cylindrical wire cage (1 m ht m diameter) wrapped in a black screen (1 mm 2 openings) with access through the top (see the Electronic supplemental material S1). Soil temperatures at 2 cm depth and air temperatures at 3 cm above the ground were measured using a metal probe thermometer within and outside of a subset of enclosures to compare the relative environmental conditions within these plots. PAR (photosynthetically active radiation) light levels within and outside the cages were also measured at 50 cm above the ground between 1200 and 1300 h during July using a LI-6400 XTi portable photosynthesis instrument (LI-COR, Inc., Lincoln, NE, USA). Cyphocleonus achates, L. minutus, or both were added to cages on the same day as they were collected from nearby field sites (Seastedt et al. 2007). They were added at densities observed in the field and similar to those used in other enclosure experiments (Smith and Mayer 2005; Corn et al. 2006). We added three male and three female C. achates to cages five weeks after transplant in August of 2006, and again during the first full growing season in July Weevils were sexed according to the abdominal shape characteristics described in Goodman et al. (2006). Adult C. achates consumed considerable amounts of rosette leaf tissue, with many leaves consumed down to the rachis during the fall of Vegetative cover hindered the monitoring of emerging adults within cages during the summer of 2007; however, 1 2 adults were noted in about half of the cages. During 2008, supplemental C. achates were not added above the existing levels. Eight adult L. minutus flower head weevils were added to each cage during the summer of the first full growing season in June of 2007 and again in 2008, at the time when this weevil normally emerges from field soils and once the majority of plants had bolted. Larinus minutus consumes vegetative and reproductive parts, primarily inflicting damage to stem, stem foliage, and flowers (Piper 2004). During both years, adult L. minutus were observed to feed on these tissues and oviposit at the base of the florets within open flowers. Plants were not artificially pollinated and flowers did not produce seeds, the primary food source for developing L. minutus larvae; thus cages did not contain flower head weevils from the previous year. We chose not to artificially pollinate in order to avoid interfering with the confinement of L. minutus and C. achates within the cages, and because our intent was to understand the effects of L. minutes herbivory on nonreproductive tissues. Our recent field study addresses the effects of these herbivores on seed production and other plant traits across a gradient of N availability and competition (Knochel and Seastedt 2010). To increase soil N availability, we added N as NH 4 NO 3 in granular form at a total rate of 38 g N m -2 year -1

5 (380 kg N ha -1 year -1 ) in three separate additions on 18 April, 16 May, and 17 June during both 2007 and Ambient levels of N availability served as control plots with no added amendments. Sucrose was used to reduce soil fertility at a rate of 252 g C m -2 year -1 in three equal 84 g C additions on the same dates as N additions. Soil amendments were sprinkled onto the soil surface and lightly watered, with control N plots also watered to control for effects of water addition. We monitored plant responses to herbivores under a wide range of soil resource conditions, from N in surplus to plant demand (perhaps representing an upper limit on availability when invading disturbed sites, pasture or agricultural lands) to low-n plots that are more representative of intact rangeland sites with greater plant competition. To assess the effects of the nutrient manipulations on soil N, soil cores were sampled cm from the target C. stoebe plants once during August Three soil cores (2 cm diameter 9 10 cm depth) were taken and composited from each of ten random plots within each of the three treatment groups (30 composited replicates). In 2008, the levels of NO 3 - and NH 4? were examined across the growing season by burying one cation anion exchange resin capsule to a depth of 10 cm from May to September within each of seven plots from the high-n and low-n treatment groups (14 replicates; each capsule contained 1 g of ionic exchange resin beads charged with H? and OH - ions held within a porous fabric membrane, UNIBEST PST-1, Bozeman, MT, USA). Resins were not buried in the control plots because resin measurements from treatment extremes were deemed adequate to provide an integrated measurement of plant-available soil N over the growing season. Upon collection, soil particles were lightly rinsed from the surface with deionized water and the resins were analyzed for N content according to the methods described for the greenhouse experiment. We recorded the number of bolting stems and the height of the tallest stem during May June In October of 2006 and 2007, C. stoebe stems and stem foliage were harvested for annual biomass. To minimize any effects of clipping on future growth and to allow the resorption of nutrients from aboveground tissues, removal of C. stoebe stems occurred after senescence, with basal rosette leaves left intact. All aboveground and root biomass was harvested in September of 2008, and we counted flowers and C. achates using the aforementioned methods. Any plants that died before terminating the experiment were similarly processed. The annual aboveground biomass of the resident plant community within each plot was also harvested by clipping at ground level in the late falls of 2006, 2007 and 2008, and processed to test for responses to fertilization, effects of neighboring plant biomass on C. stoebe growth, or effects of C. stoebe on the growth of neighbors as a function of the herbivore and soil nutrient treatments. Statistical analysis Within individual growing seasons, the effects of soil amendments on soil KCl extractable inorganic N and plant traits and the effects of insects on plant traits were analyzed using analysis of variance (ANOVA) procedures, and treatment differences were analyzed using a posteriori Ryan Einot Gabriel Welsch multiple-range tests (SAS v.9.2). Across the two growing seasons ( ), SAS PROC MIXED repeated-measures ANCOVA was used to assess the effects of C. achates, L. minutus, or their interaction, as well as soil treatments and their interaction with insect treatments, on spotted knapweed aboveground biomass and flower production. We also assessed neighboring plant aboveground biomass responses to soil amendments from 2006 to A priori hypotheses for the effects of insect (using both species vs. no insects, or one insect versus no insects) and soil N were tested using single degree-of-freedom contrast statements. Insect or soil treatments and their interaction were the between-subject fixed factors, year was the repeated within-subject factor, and the initial plant size was used as a random covariate. We used a nonsignificant interaction term between insect treatments in these analyses to indicate that damage from the two species was additive. Values for Akaike s information criterion (AIC) and Schwarz s Bayesian criterion (SBC) guided the choice of unstructured as the best covariance structure for these repeated-measures data. To assess the potential effect of plant neighbors on C. stoebe performance, we used a stepwise linear regression analysis with the total biomass of individual target C. stoebe plants in 2007 or 2008 as the dependent variable and neighboring biomass (g m -2 ) within individual enclosures in the year 2006, 2007, or 2008 as explanatory variables. A significance level of P = 0.05 was used for the inclusion and retention of parameters in the model. Any response variables that did not meet equality of variance assumptions were transformed and then back-transformed for use in figures. All dead C. stoebe plants in 2008 were included in statistical analyses, with values of dead root or stem biomass recorded or entered as zero if the plants produced no aboveground tissue [log transformations (using x? 1) of data with zeros were sufficient to meet assumptions of normality]. Percent mortality was compared among insect or nutrient treatments using a chi-square analysis. Results Greenhouse experiment Taproots of C. stoebe contained between zero and nine C. achates (0 6 in the reduced-n group), with 75% having

6 between one and five, and an average of 2.36 ± 0.26 individuals per root. This gradient of larval densities was equivalent to that observed at a biological control release site in Colorado (Seastedt et al. 2007). Plants from control and low-n soils did not differ statistically in their C. achates density (control mean = 2.17 ± 0.36 C. achates/root, low-n mean = 2.58 ± 0.37 C. achates/ root, F 1,55 = 0.63, P = 0.43), nor in developmental stage. Carbon amendments resulted in a decrease in inorganic N below the levels observed for control soils (F 1,7 = 7.56, P = 0.033). Plants grown in the low-n soils produced less total biomass (8.2 ± 0.3 g) than those in control soils (10.4 ± 0.3 g) (F 1,51 = 19.66, P \ ). Neither the root-to-shoot ratio (F 1,51 = 0.53, P = 0.47) nor flower production (F 1,55 = 1.09, P = 0.30) differed between soil treatments. Four plants in the greenhouse died during bolting and before flower production, representing a 7% mortality rate. All dead plants were in the low-n treatment group, and contained an average of 3.0 C. achates weevils (range of 1 5). Weevils appeared to induce mortality by consuming most of the root tissue. Thus, root biomass and flower production for these plants were recorded as zero. In an ANCOVA, soil fertility and density of C. achates had significant effects on root biomass (soil fertility, F 1,51 = 3.99, P = 0.051, C. achates density F 1,52 = 5.91, P = 0.019) and also on total biomass (soil fertility, F 1,51 = 11.97, P = ; C. achates density F 1,51 = 7.27, P = ) (Fig. 1). Shoot biomass also showed no significant trends in relation to C. achates density in any soil treatment (data not shown). A model including the effects of soil treatment, weevil density, and their interaction on flower production was significant (F 4,52 = 7.19, P \ 0.001, R 2 = 0.38). Flower production showed modest decreases with increasing C. achates density while controlling for shoot biomass (F 1,52 = 7.36, P = 0.009) (Fig. 2). A regression that omitted the root sample containing nine weevils also remained significant, P = The soil N by C. achates density interaction was not significant (F 1,52 = 1.75, P = 0.19). Common garden experiment Environmental conditions Soil samples in 2007 and resin bags in 2008 both confirmed that soil treatments had the intended effects on KCl-extractable inorganic N. In 2007, high-n plots had significantly greater (164.9 ± 40.7 lg Ng -1 soil) and C-amended soils significantly lower (1.46 ± 0.47 lg Ng -1 soil) levels of N compared with control plots (3.89 ± 0.54 lg Ng -1 soil; F 2,20 = 46.13, P \ ). Similarly, resin-captured N in 2008 was significantly higher in the a 12 Reduced N R 2 = 0.27 = Total biomass p = = Root biomass Biomass (g plant -1 ) b Biomass (g plant -1 ) Ambient N R 2 = 0.24 p = Cyphocleonus density (No. root -1 ) Fig. 1 Relationship between root weevil density and root and total biomass in a reduced (low-n) soil (root: F 1,28 = 6.13; total: Pearson s r = 0.52, F 1,20 = 7.02) and b control soil (root: F 1,28 = 1.27; total: Pearson s r = 0.22, F 1,28 = 1.38) Flowers per plant R 2 = 0.08 p = Cyphocleonus density (No. root -1 ) Fig. 2 Relationship between flower production and abundance of C. achates fertilized plots (152.8 ± 42.3 lg Ng -1 resin) versus low-n plots (2.49 ± 0.93 lg Ng -1 resin; F 1,13 = 51.12, P \ ). Gravimetric soil moisture in the 2007 samples did

7 not differ between soil treatments (F 1,20 = 0.36, P = 0.71). Soil temperatures at 2 cm depth were approximately 2 C lower on the inside versus the outside of cages, and air temperatures were about 6 C lower; however, the temperatures for different cages or for different insect and soil treatments were equivalent. Light intensity (PAR) at midday inside the screened cages was 1,150 lm m -2 s -1, a reduction of 42% compared to full sunlight (1,967 lm m -2 s -1 ), but still over eight times greater than the light intensity in the shade of an adjacent deciduous forest canopy (138 lm m -2 s -1 ). Soil nutrient effects on C. stoebe and neighboring plant biomass Centaurea stoebe stem biomass was only significantly different between soil treatments in 2008 (Tables 1, 2). Analyzing cumulative results over two seasons, plants growing in the control soil treatment group had a higher mean flower production than those in the high-n plots (Table 1). Responses of plant neighbors to soil treatments contrasted with those of C. stoebe. While spotted knapweed biomass was higher in control plots, the collective biomass of the plants neighboring each target C. stoebe individual during was greater in high-n versus control and high-n versus low-n plots (Tables 1, 2). The mean neighboring plant biomass across all treatments was ± g m -2 in 2007 and ± 8.68 g m -2 in A negative relationship was detected between total C. stoebe biomass in 2008 and the aboveground biomass of neighboring plants measured during the two previous years (2006 and 2007) (F 2,139 = 5.48, P = ). Neighboring plant biomass in 2006 and 2007 explained 7% of the total variation in C. stoebe biomass in When analyzed by soil treatment, total C. stoebe biomass was negatively correlated with neighboring plant biomass in high-n plots (F 2,44 = 4.07, P = 0.024, R 2 = 0.16), but exhibited no such relationship in the control or low-n treatments. Cyphocleonus achates abundance For caged plots, plant roots in either insect treatment that included C. achates (n = 71) contained an average of 0.74 ± 0.13 individuals per root, and ranged from zero to four C. achates, with 41% of roots (29 plants) infested (larvae, pupae, adults or chambers). During the late summer and fall of 2006, rosettes subjected to the C. achates treatment sustained a 34% decrease in maximum diameter when compared to control cages. Plants within different soil treatments did not differ in the number of C. achates per root, and C. achates density was not significantly correlated with initial plant size or final root mass (data not shown). Effects of herbivory on plant performance Over two growing seasons, 23% (32) of all experimental plants died (12 in 2007 and 20 in 2008). Of the plants that died, 34% had been treated with both C. achates and L. minutus herbivory, 28% had been treated with C. achates only, 22% were L. minutus-only plots, and 16% died in the absence of either insect, perhaps representing a background mortality rate. There was a marginally significant difference in percent C. stoebe mortality among the insect treatments [v 2 (df = 3, N = 142) = 7.2, P = 0.066]. There was a significant difference in percent C. stoebe mortality between the nutrient treatments [v 2 (df = 2, N = 142) = 33.1, P \ 0.001]. The high-n group represented 59% of the dead plants, followed by the low-n group (28%), with the least mortality associated with plants with control soil N (13%). All of the caged plants had bolted by 16 May 2007, and on that date the plants grew significantly more stems in cages with no insects (mean = 5.00 ± 0.58 stems plant -1 ) compared to those with C. achates, L. minutus, or both insects combined (mean B 3.33 ± 0.44 stems) (F 3,141 = 3.63, P = 0.014). On 17 June 2007, plants were also significantly taller in the cages with no insects (mean = ± 2.20 cm) or L. minutus only (mean = ± 2.93 cm) compared to those plants with C. achates (mean = ± 2.94 cm) or C. achates? L. minutus (mean = ± 2.65) (F 3,141 = 7.88, P \ ). Repeated-measures factorial analysis of the main effect of insects on aboveground biomass and flower production over two consecutive field seasons showed no statistical interaction between C. achates and L. minutus (F 1,135 = 0.74, P = 0.392). However, biomass and flowering were significantly lower (approximately half) in the insect treatment with both C. achates and L. minutus than for those plants with no biological control insects (Tables 1, 2; Fig. 3). A priori contrasts showed that plants treated with both C. achates and L. minutus produced half the biomass of those with no insects (F 1,135 = 24.3, P \ ), and about a third less biomass than plants with only L. minutus (F 1,135 = 9.3, P = ). These analyses indicated that a summation of the individual negative effects of the two species was equivalent to the negative effects in plots where both were added. There was no significant difference in mean aboveground biomass between plants with only L. minutus or C. achates versus those with no insects. Flower production followed a similar pattern: plants with the combination of insect species produced fewer flowers on average than plants without insects (F 1,135 = 6.40, P = ; F 1,135 = 4.41, P = , respectively), but L. minutus and C. achates alone did not reduce flowering compared to control plots with no insects. We

8 Table 1 Effects of soil and insect treatments in the common garden experiment on C. stoebe response variables in 2007, 2008, and over the entire experiment ( ) using GLM and mixed-model repeated-measures ANCOVA Measurement Soil N Insects High Control Low C. achates? L. minutus C. achates L. minutus None (a) 2007 Stem biomass (g) a 22.7 ± 5.2 a 18.8 ± 2.7 a 15.9 ± 2.3 a 12.2 ± 2.4 a 16.4 ± 3.3 ab 18.4 ± 2.9 bc 29.7 ± 6.5 c Flowers/plant 65.1 ± 15.0 a 84.3 ± 17.5 a 67.2 ± 11.8 a 53.2 ± 12.6 a 65.9 ± 15.3 a 67.6 ± 16.3 a 102 ± 22.6 b (b) 2008 Total biomass (g) 21.8 ± 5.4 a 44.0 ± 7.7 b 43.8 ± 8.2 b 23.6 ± 6.8 a 46.6 ± 9.2 ab 36.7 ± 8.8 ab 40.1 ± 8.8 b Root:shoot ratio 0.46 ± 0.09 a 0.28 ± 0.02 b 0.32 ± 0.04 b 0.36 ± 0.04 a 0.29 ± 0.06 a 0.35 ± 0.05 a 0.36 ± 0.08 a Root biomass (g) 4.08 ± 0.7 a 6.8 ± 0.9 b 6.4 ± 1.0 ab 3.97 ± 0.8 a 6.5 ± 1.1 ab 5.6 ± 0.99 ab 7.1 ± 1.1 b Aboveground biomass (g) 17.7 ± 4.8 a 37.2 ± 6.8 b 37.3 ± 7.3 b 19.6 ± 6.1 a 40.1 ± 8.2 b 31.1 ± 7.9 ab 32.9 ± 7.7 c Flowers/plant 31.2 ± 8.6 a 72.0 ± 11.9 b 60.6 ± 11.8 b 31.4 ± 9.7 a 73.2 ± 14.8 b 55.8 ± 13.6 b 57.9 ± 12.1 b (c) Aboveground biomass (g) 20.2 ± 3.5 a 28.0 ± 3.8 a 26.6 ± 4.0 a 15.9 ± 3.3 a 28.3 ± 4.6 ab 24.8 ± 4.2 ab 31.3 ± 5.0 b Flowers/plant 47.9 ± 8.8 a 77.4 ± 10.5 b 66.5 ± 8.6 ab 45.8 ± 8.7 a 69.6 ± 10.6 ab 61.1 ± 10.5 ab 80.2 ± 13.0 b Neighbor biomass (g/m 2 ) ± 7.1 a ± 5.0 b ± 4.2 b Different letters denote significant differences among soil or insect treatments; P B Values are mean ± 1SE a 2007 stem biomass excludes rosette foliage, which was left intact until full aboveground tissue harvest in 2008 Table 2 ANCOVA results for the common garden experiment for 2007 and 2008, and mixed-model repeated-measures ANCOVA results over both seasons, relating to the effects of nutrient and insect treatments and interactions of them on plant response variables Variables Soil N Insects Initial rosette size Soil N 9 insects df F P df F P df F P df F P (a) 2007 Stem biomass a 2, , < , < , Flower production 2, , , < , (b) 2008 Total biomass 2, , , < , Root:shoot ratio 2, , , , Root biomass 2, , , < , Aboveground biomass 2, < , , < , Flowers/plant 2, < , , < , (c) Aboveground biomass 2, , < , < , Flowers/plant 2, , , < , Neighbor biomass 2, <0.001 P values in bold denote significant differences among treatments a 2007 stem biomass excludes rosette foliage, which was left intact until full aboveground tissue harvest in 2008 found no significant insect by soil fertility interactions, such that the negative effects of herbivory on biomass and flower production did not depend on soil N level. However, a posteriori comparisons over the two-season period revealed statistically significant trends between insect treatments for plants in two of the three soil conditions. Both within and among years, insect treatments were associated with greater variation in biomass within the high-n and control plots than within low-n plots, with significant differences in biomass noted between both-insect and no-insect treatments during (F 1,130 = 4.04, P \ 0.05). Discussion Spotted knapweed has successfully invaded areas with historically low nitrogen availabilities (Hooper and

9 a Flowers per plant Aboveground biomass (g plant -1 ) b A A C. achates C. achates L. minutus None + L. minutus Fig. 3 Effects of treatment on a aboveground biomass and b flower production over two years. Points are mean ± SE (n = 141). Different letters indicate statistically significant differences based on a priori contrasts between treatment means, P B 0.05 Johnson 1999), but it is also an opportunistic species that can benefit from high resource conditions and disturbance (Story et al. 1989; Knochel and Seastedt 2010; Knochel et al. 2010). However, taken together, our experiments suggest that the discrepancies among reports of the impact of spotted knapweed biological control are best explained by the intensity of plant competition and the numbers and species of herbivores present. In the greenhouse study, we found that flower production was reduced in response to increasing intensity of root herbivory by C. achates per unit of stem biomass, and this negative effect did not interact with soil resource conditions (Fig. 2). The decrease in flowering was concomitant with an increase in root biomass (Fig. 1), and an increase in N allocated to the root tissues, both of which are potentially a consequence of gall-like tissue causing a thickening around the areas damaged by weevils. The purpose of this study was not to elucidate specific physiological mechanisms of plant responses to herbivory under varying resource regimes, which are complex (Trumble et al. 1993; Stamp 2003). However, the observed increase in root N content with increased weevil density supports the idea that roots may have increased N acquisition to support plant needs and AB AB AB AB B B compensate for root damage. Meanwhile, we observed no significant changes to aboveground biomass under more intense root herbivory. These results contrast with those showing C. stoebe overcompensation responses to Agapeta zoegana larval herbivory, as reported by Ridenour and Callaway (2003) and Newingham et al. (2007). Combined with these studies, our work reiterates that compensation responses can be highly dependent on the species of root herbivore and the intensity of damage. Agapeta zoegana and C. achates likely caused different levels of compensation by spotted knapweed because each herbivore damages structurally distinct tissues within the root (Müller 1989), with C. achates development in the central vascular tissue apparently being more detrimental than Agapeta damage to the root cortex. Although we found hints of a slight inclination toward an interaction effect between root herbivory and soil N on flower production, our results do not support the hypothesis that root herbivory interacts with soil N availability, or that herbivory reduces plant fitness more dramatically under low-nutrient conditions. Thus, as a variation on the compensatory continuum model (Maschinski and Whitham 1989), impacts of C. achates on individual plants may be influenced more by insect density and damage intensity than by soil N availability. Nonetheless, the greenhouse experiment did not include plant competition, and results from the common garden experiment suggest that fitness declines could be more severe in the presence of plants competing for a shared soil resource (Müller-Schärer 1991). Effects of multiple herbivores under varied abiotic conditions Overall, we observed additive negative effects of simultaneous above- and belowground herbivory by C. achates and L. minutus weevils in the common garden study relative to controls. This combination of insects reduced flowering and biomass over a two-year period, as predicted by the magnitude of individual insect effects (Fig. 3; Table 1). These results fit predictions of the cumulative stress hypothesis (Müller-Schärer and Schroeder 1993; McEvoy and Coombs 1999), which states that a combination of insects may have a more detrimental impact on individual plants, and therefore has the potential to improve biological control if this impact scales up to negative effects on population densities. The presence of either insect species was never found to result in a net increase (or overcompensation response) in the measured indices of C. stoebe fitness (Fig. 3; Table 2a c). Cyphocleonus achates larval densities and the negative effects of root damage on flowering in the greenhouse experiment were remarkably similar to the densities and effects on plants over multiple years at a nearby field

10 release site (Knochel and Seastedt 2010). In comparison, larval numbers were above the mean of 0.3 larvae per root surveyed by Sturdevant et al. (2006), but were substantially below the means of 11.9 and 13.5 larvae per root reported over two consecutive years in a manipulative experiment by Corn et al. (2007). Thus, we believe that the levels of C. achates damage and the effects on flowering that we observed are well within the realistic range found under field conditions, if not conservative estimates. Interestingly, although larval root densities of C. achates in the caged experiment were much lower than those observed in the greenhouse or in a companion field study (Knochel and Seastedt 2010), and were more similar to those reported across multiple field sites by Sturdevant et al. (2006), the weevils nonetheless reduced plant growth and flower production, especially when combined with L. minutus. Beyond root damage, the consumption of young rosette and foliage tissues by C. achates during 2006 and 2007 appears to have had an enduring negative effect on plant fitness. This fits the prediction by Myers and Risley (2000) that reducing rosette survival or growth may be a key part of the successful biological control of some invasive Centaurea species. Further, similar to Ridenour et al. (2008) and Corn et al. (2007), who reported mortalities of 66 and 13% due to C. achates during their experiments, we also found that a substantial proportion of the rosettes and bolting plants died prematurely, with 28% mortality observed in cages containing only C. achates. The influence of C. achates herbivory on mortality relative to other factors is unknown; however, as a group, these observations indicate that C. achates can induce mortality under some conditions. The few significant differences in plant traits among soil N treatments were largely inconsistent. Unexpectedly, during 2008, the total spotted knapweed biomass and the flower production were 50% lower in the high-n plots compared to control and low-n plots (Table 2b). In contrast, neighboring grasses and forbs within the same plots greatly increased in biomass over the experiment in response to N addition. For unknown reasons, the neighboring plant community, dominated by the Eurasian grass Agropyron cristatum, appeared to have gained the competitive advantage and may have caused water or other limitations on C. stoebe. This result is analogous to the increase in the competitive strength of Bromus inermis over C. stoebe in higher-n soils (Lindquist et al. 1996). We observed a similar competitive disadvantage for seedlingstage C. stoebe plants, where grass competition under elevated N conditions greatly reduced seedling establishment and biomass (Knochel et al. 2010). Our studies support the contention that declines in seed production and spotted knapweed densities over several years and in multiple locations across North America have resulted not only from direct seed reductions by L. minutus or Urophora spp. (Story et al. 2008) but also the negative effects of L. minutus and C. achates feeding on above- and belowground tissues. We exercise caution in scaling up our relatively short-term results from individual plants to longterm trends at the level of a field population, but we believe that the evidence reported here and from recently published field studies (Knochel and Seastedt 2010; Knochel et al. 2010) substantially improve understanding of above- and belowground herbivory in concert, and provide a probable explanation for some of the documented C. stoebe population declines across North America. This evidence, in addition to studies reporting on the capacity of existing vegetation to resist invasion by spotted knapweed at realistic propagule densities (e.g., Pokorny et al. 2005; Knochel et al. 2010), substantiate the hypothesis that herbivory and plant competition are capable of greatly reducing the establishment and spread of this invasive species. Our general findings of negative impacts of herbivores corroborate with the observation of Story et al. (2008) that these insects can have strong negative effects on C. stoebe densities. The cage experiment represented the first study to examine the single and combined effects of above- and belowground herbivores on spotted knapweed, and further, to monitor these effects across a gradient in resource availability. Perhaps most importantly, we found that C. achates and L. minutus in combination had greater negative effects on plant performance than either species alone. Thus, the inconsistent reports of biological control success on spotted knapweed may be due in part to differences in the damage intensity and types of insect species present. The potential to utilize biocontrol insects to reduce spotted knapweed dominance and effectively manage the plant at a regional scale will likely depend on the presence of multiple species of herbivores and efforts to increase competition with desirable native vegetation. Further, long time periods are necessary to see the negative effects of insect activities on plant densities (Seastedt et al. 2007; Story et al. 2008; Knochel and Seastedt 2009). However, we expect the combination of below- and aboveground herbivory and plant competition to provide the cumulative stress needed to diminish the dominance of the aggressive invasive species C. stoebe in North America. Acknowledgments We thank Christine Fairbanks, Kali Blevins, and Tate Seastedt for help with the greenhouse and cage experiments. Drs. Deane Bowers, William Bowman, Carol Wessman, Susan Beatty, and three anonymous reviewers provided valuable input on earlier drafts of the manuscript. Dr. Thomas Lemieux and the University of Colorado graciously allowed us to convert a grassland into a weed garden. This work has been funded by the National Research Initiative of the USDA Cooperative State Research, Education and Extension Service, grant number All experiments comply with current US laws.

11 References Blumenthal DM, Jordan NR, Russelle MP (2003) Soil carbon addition controls weeds and facilitates prairie restoration. Ecol Appl 13: Coombs EM, Clark JK, Piper GL, Cofrancesco AF Jr (2004) Biological control of invasive plants in the United States. Oregon State University Press, Corvallis Corn JG, Story J, White LJ (2006) Impacts of the biological control agent C. achates on spotted knapweed, Centaurea maculosa, in experimental plots. Biol Control 37:75 81 Corn JG, Story JM, White LJ (2007) Effect of summer drought relief on the impact of the root weevil Cyphocleonus achates on spotted knapweed. Environ Entomol 36: Corn JG, Story JM, White LJ (2009) Comparison of larval development and overwintering stages of the spotted knapweed biological control agents Agapeta zoegana (Lepidoptera: Tortricidae) and Cyphocleonus achates (Coleoptera: Curculionidae) in Montana versus eastern Europe. Environ Entomol 38: Cortilet, AB, Northrop N (2006) Biological control of European buckthorn and spotted knapweed (Minnesota Dept Agriculture Final Program Report). Minnesota Dept Agriculture, St Paul (see weedcontrol/knapweedlcmrfinalreport.pdf) Duncan CA, Jachetta JJ, Brown ML, Carrithers VF, Clark JK, DiTomaso JM, Lym RG, McDaniel KC, Renz MJ, Rice PM (2004) Assessing the economic, environmental, and societal losses from invasive plants on rangeland and wildlands. Weed Technol 18: Goodman CL, Phipps SJ, Wagner RM, Peters P, Wright MK, Nabli H, Saathoff S, Vickers B, Grasela JJ, McIntosh AH (2006) Growth and development of the knapweed root weevil, Cyphocleonus achates, on a meridic larval diet. Biol Control 36: Hill JP, Germino MJ (2005) Coordinated variation in ecophysiological properties among life stages and tissue types in an invasive perennial forb of semiarid shrub steppe. Can J Bot 83: Hill JP, Germino MJ, Wraith JM, Olsen BE, Swan MB (2006) Advantages in water relations contribute to greater photosynthesis in Centaurea maculosa compared with established grasses. Int J Plant Sci 167: Hooper DU, Johnson L (1999) Nitrogen limitation in dryland ecosystems: responses to geographical and temporal variation in precipitation. Biogeochemistry 46: Hufbauer RA, Sforza R (2008) Multiple introductions of two invasive Centaurea taxa inferred from cpdna haplotypes. Divers Distrib 14: Jacobs JS, Sing SE, Martin JM (2006) Influence of herbivory and competition on invasive weed fitness: observed effects of Cyphocleonus achates (Coleoptera: Curculionidae) and grassseeding treatments on spotted knapweed performance. Environ Entomol 35: Keeney DR, Nelson DW (1982) Nitrogen inorganic forms. In: Page AL et al (eds) Methods of soil analysis (Agronomy Monograph 9, part 2), 2nd edn. American Society of Agronomy, Madison, pp Kennett GA, Lacey JR, Butt CA, Olson-Rutz KM, Haferkamp MR (1992) Effects of defoliation, shading and competition on spotted knapweed and bluebunch wheatgrass. J Range Manag 45: Knochel DG, Seastedt TR (2009) Sustainable control of spotted knapweed, (Centaurea stoebe). In: Inderjit (ed) Management of invasive weeds. Springer, Berlin, pp Knochel DG, Seastedt TR (2010) Reconciling contradictory findings of herbivore impacts on the growth and reproduction of spotted knapweed (Centaurea stoebe). Ecol Appl 20(7): Knochel DG, Flagg C, Seastedt TR (2010) Effects of plant competition, seed predation, and nutrient limitation on seedling survivorship of spotted knapweed (Centaurea stoebe). Biol Invasions (in press) Lacey JR (1989) Influence of spotted knapweed (Centaurea maculosa) on surface runoff and sediment yield. Weed Tech 3: Lindquist JL, Maxwell BD, Weaver T (1996) Potential for controlling the spread of Centaurea maculosa with grass competition. Great Basin Naturalist 56: Maron JL (1998) Insect herbivory above- and belowground: individual and joint effects on plant fitness. Ecology 79: Maschinski J, Whitham TG (1989) The continuum of plant responses to herbivory: the influence of plant association, nutrient availability, and timing. Am Nat 134:1 19 Masters GJ, Brown VK, Gange AC (1993) Plant mediated interactions between aboveground and belowground insect herbivores. Oikos 66: Mattson WJ (1980) Herbivory in relation to plant nitrogen content. Annu Rev Ecol Syst 11: McEvoy PB, Coombs EM (1999) Biological control of plant invaders: regional patterns, field experiments, and structured population models. Ecol Appl 9: Michels GJ Jr, Carney VA, Jurovich D, Kassymzhanova-Mirik S, Jones E, Barfot K, Bible J, Mirik M, Best S, Bustos E, Karl B, Jimenez D (2009) Biological control of noxious weeds on federal installations in Colorado and Wyoming (Texas Agricultural Experiment Station 2009 Annual Report). Texas Agricultural Experiment Station, College Station (see edu/programs/entotaes/noxiousweeds/coloradonoxiousweeds. html) Müller H (1989) Growth pattern of diploid and tetraploid spotted knapweed, Centaurea maculosa Lam. (Compositae), and effects of the root-mining moth Agapeta zoegana (L.) (Lep.: Cochylidae). Weed Res 29: Müller-Schärer H (1991) The impact of root herbivory as a function of plant density and competition: survival, growth, and fecundity of Centaurea maculosa in field plots. J Ecol 28: Müller-Schärer H, Schroeder D (1993) The biological control of Centaurea spp. in North-America do insects solve the problem? Pestic Sci 37: Myers JH, Risley C (2000) Why reduced seed production is not necessarily translated into successful biological weed control. In: Spencer NR (ed) Proceedings of the X International Symposium on the Biological Control of Weeds. Montana State University, Bozeman, pp Myers JH, Jackson C, Quinn H, White S, Cory JS (2009) Successful biological control of diffuse knapweed, Centaurea diffusa, in British Columbia, Canada. Biol Control 50:66 72 Newingham BA, Callaway RM, BassiriRad H (2007) Allocating nitrogen away from an herbivore: a novel compensatory response to root herbivory. Oecologia 153: NOAA (2009) Earth System Research Laboratory (Physical Sciences Division) website. National Oceanic and Atmospheric Administration, Washington, DC (Montana: psd/data/timeseries/; Colorado: Boulder.mm.precip.html) NRCS (2009) National Resources Conservation Service (Web Soil Survey) website. United States Department of Agriculture, Washington, DC (see Ochsmann J (2001) On the taxonomy of spotted knapweed. In: Smith L (ed) Proceedings of the First International Knapweed Symposium of the 21st Century, Coeur d Alene, ID, USA, March 2001, pp Pearson DE, Fletcher RJ (2008) Mitigating exotic impacts: restoring deer mouse populations elevated by an exotic food subsidy. Ecol Appl 18: