Shifts in dynamic regime of an invasive lady beetle are linked to the invasion and insecticidal management of its prey

Transcription

1 Ecological Applications, 25(7), 2015, pp Ó 2015 by the Ecological Society of America Shifts in dynamic regime of an invasive lady beetle are linked to the invasion and insecticidal management of its prey CHRISTINE A. BAHLAI, 1,5 WOPKE VANDER WERF, 2 MATTHEW O NEAL, 3 LIA HEMERIK, 4 AND DOUGLAS A. LANDIS 1 1 Department of Entomology, Michigan State University, East Lansing, Michigan, USA 2 Crop Systems Analysis Group, Department of Plant Sciences, Wageningen University, Wageningen, The Netherlands 3 Department of Entomology, Iowa State University, Ames, Iowa, USA 4 Biometris, Department of Mathematical and Statistical Methods, Wageningen University, Wageningen, The Netherlands Abstract. The spread and impact of invasive species may vary over time in relation to changes in the species itself, the biological community of which it is part, or external controls on the system. We investigate whether there have been changes in dynamic regimes over the last 20 years of two invasive species in the midwestern United States, the multicolored Asian lady beetle Harmonia axyridis and the soybean aphid Aphis glycines. We show by model selection that after its 1993 invasion into the American Midwest, the year-to-year population dynamics of H. axyridis were initially governed by a logistic rule supporting gradual rise to a stable carrying capacity. After invasion of the soybean aphid in 2000, food resources at the landscape level became abundant, supporting a higher year-to-year growth rate and a higher but unstable carrying capacity, with two-year cycles in both aphid and lady beetle abundance as a consequence. During , farmers in the Midwest progressively increased their use of insecticides for managing A. glycines, combining prophylactic seed treatment with curative spraying based on thresholds. This human intervention dramatically reduced the soybean aphid as a major food resource for H. axyridis at landscape level and corresponded to a reverse shift towards the original logistic rule for year-to-year dynamics. Thus, we document a short episode of major predator prey fluctuations in an important agricultural system resulting from two biological invasions that were apparently damped by widespread insecticide use. Recent advances in development of plant resistance to A. glycines in soybeans may mitigate the need for pesticidal control and achieve the same stabilization of pest and predator populations at lower cost and environmental burden. Key words: Aphis glycines; area-wide management; Coccinellidae; Harmonia axyridis; multitrophic interaction; neonicotinoid pesticides; predator prey relationships; regime shift; Ricker model. INTRODUCTION Regime shifts are substantial, sudden, and persistent changes in the function of an ecosystem or process (Scheffer et al. 2001, Carpenter and Brock 2006, Carpenter et al. 2008, Biggs et al. 2009) where a regime is a characteristic behavior of a system which is maintained over time through feedbacks within the system. Regime shifts can be brought about by single disturbances shocking the system into a new state or by smooth changes to a driver of one of the system s internal processes until some threshold is surpassed (Scheffer and Carpenter 2003, Folke et al. 2004). Transitions between states in food webs represent an important class of regime shifts (Ives and Carpenter 2007, Carpenter et al. 2008). Prey dynamics can dramatically shift the population dynamics of predatory species (Persson et al. 2007). Thus, novel, invading prey may cause notable shifts in dynamic regimes governing Manuscript received 3 November 2014; revised 12 February 2015; accepted 5 March Corresponding Editor: M. P. Ayers. 5 cbahlai@msu.edu predator population regulation. Causal mechanisms for these large changes in population dynamics may include overexploitation of resources, an increase in the impact of natural enemies, genetic and evolutionary shifts, or habitat changes (Strauss et al. 2006, Strayer et al. 2006). In agricultural systems, invasive herbivores are widespread and threaten productivity of crop plants (Vitousek et al. 1996, Mack et al. 2000, Hulme 2009). Invasive predators may also indirectly affect plant productivity through cascading effects caused by prey consumption or interference with native predator guilds (Kenis et al. 2009, Snyder 2009, Crowder and Snyder 2010). In this study, we focus on the population interactions of multicolored Asian lady beetle (Harmonia axyridis) and soybean aphid (Aphis glycines), two invasive species in an agricultural ecosystem of the American Midwest. Harmonia s invasion, costs and benefits Harmonia was first detected in the north central United States in (Koch et al. 2006) and rapidly became one of the most abundant coccinellid species in arable crop landscapes (Colunga-Garcia et 1807

2 1808 CHRISTINE A. BAHLAI ET AL. Ecological Applications Vol. 25, No. 7 al. 1997, Colunga-Garcia and Gage 1998, Bahlai et al. 2013, 2015). Harmonia s invasion is associated with a long list of negative ecological, economic, and quality of life impacts. Due to its relatively large size, voracity, and competitive ability, Harmonia has been associated with the decline of native lady beetles (Harmon et al. 2007, Roy et al. 2012). Harmonia is an economic pest of wine grapes (Pickering et al. 2004, 2005), and it is also considered a nuisance pest of humans because of its biting behavior, its habit of entering homes in the fall, and as a cause of human allergies (Koch 2003, Galvan et al. 2007, Nakazawa et al. 2007, Koch and Galvan 2008). However, Harmonia is also widely considered a beneficial insect because it contributes to the biological control of a wide variety of pests, particularly aphids (McClure 1987, Brown and Miller 1998, Koch 2003, Specty et al. 2003, Fox et al. 2004, Costamagna and Landis 2006, Pervez and Omkar 2006, Costamagna and Landis 2007). Soybean aphid and its natural enemies Soybean aphid is considered the single most important pest of soybeans in North America (Ragsdale et al. 2011). Originating in Asia (Wu et al. 2004), this species was initially discovered in North America in 2000 (Ragsdale et al. 2004). Immediately following its arrival, soybean aphid displayed a striking pattern of alternating outbreak and non-outbreak years across wide areas of the Midwestern USA (Rhainds et al. 2010). The twoyear cycle has been attributed to the action of an effective natural enemy community and especially to Harmonia as the dominant predator of soybean aphid (Bahlai and Sears 2009, Heimpel et al. 2010, Rhainds et al. 2010). Harmonia populations are observed to have an apparent delayed density dependence associated with soybean aphid: the lady beetle reaches very high numbers in autumns following summers when soybean aphid is abundant (Bahlai and Sears 2009), and large numbers of Harmonia are captured in the following growing season, when soybean aphid numbers are low (Knapp et al. 2012). When infestations are severe, populations of soybean aphid can reach tens of thousands of aphids per plant and result in up to 40% yield losses (DiFonzo and Hines 2002). Formerly a crop with very low insecticide use, the occurrence of such outbreaks prompted widespread interest in the use of insecticides in soybean to control its damage. Effective insecticide products were rapidly identified and economic injury levels were developed (Ragsdale et al. 2011, Tilmon et al. 2011). Now, insecticides are widely used to control this aphid. However, there has been a general decline in soybean aphid populations since they last peaked in 2005, and no major outbreaks have occurred across the Midwestern United States since Since then, outbreaks have not occurred on regional scales and reports of local infestations requiring foliar-applied insecticide treatment are rare. Study objectives We examine the dynamics of a local population of Harmonia over two decades after its initial invasion in the North American Midwest and assess evidence for changes in dynamic regime by model selection (Hilborn and Mangel 1997, Bolker 2008). We evaluate changes observed relative to changes in patterns of soybean aphid outbreaks in the greater Midwest region (specifically, Michigan, Wisconsin, Iowa, and Illinois). We then examine the available evidence for potential drivers of these changes. METHODS Harmonia population dynamics Yearly average per-trap captures of Harmonia axyridis from 1994 to 2013 from Michigan State University s Kellogg Biological Station (KBS) were used to model population regulation processes in this species (available online). 6 Data were collected at KBS in southwestern Michigan ( N, W) at 288 m elevation as described in Bahlai et al. (2013). Coccinellid surveillance data were collected starting in 1989 at the KBS Long Term Ecological Research (LTER) site as part of the Main Cropping System Experiment (MCSE) and monitoring of nearby forest sites (Landis and Gage 2015). The MCSE consists of seven treatments, including threeyear rotations of annual field crops (maize, soybeans, and wheat) under four levels of management intensity (conventional, no-till, reduced input, and biologically based), as well as three continuous cultivation systems: alfalfa, poplar, and early successional vegetation (i.e., abandoned agricultural fields). Each treatment is replicated six times with plot sizes of 1 ha. Sticky card samples are taken from five permanent sampling stations within each plot. Coccinellid sampling in nearby forests began in 1993 (all sites within 3 km of the MCSE on the KBS). Forest types included year old conifer forest plantations, late successional deciduous forest, and successional forests occurring on abandoned agricultural land. Three 1-ha replicates of each forest type were monitored at five sampling stations per replicate, as in the MCSE. To monitor coccinellids, unbaited two-sided, yellow cardboard sticky cards (Pherocon, Zoecon, Palo Alto, California, USA) were suspended at 1.2 m aboveground at each sampling station. Traps were changed weekly over the growing season, usually late May to mid-september at KBS, but varied with crop sampled, availability of labor, and weather. These data are available in the KBS data archive (see footnote 6). To calculate average density, the total number of Harmonia captured in a sampling year at KBS was divided by the total number of exposed traps (a total of 300 trap locations over 10 treatments, six replicates per treatment, and five sampling stations per replicate, with a sequence of weekly samples at each station). The beginning of each 6

3 October 2015 REGIME SHIFTS IN AN INVASIVE PREDATOR 1809 sampling period was in mid-may. However, the length of the sampling period varied from year to year, and data were therefore culled at 28 August (day of the year 240) to minimize an effect of sampling period on the results. With these final sample data, the length of the yearly sampling periods ranged from 8 to 19 weeks, with a mean sampling period of 13 weeks. To determine if a shift in dynamic regime occurred during the study period and to identify break points in the time series, if present, an iterative model-selection approach was used. First, a density dependent model was fitted to the entire time series of lady beetle abundance values Next, the time series was subdivided into two to three subsets of consecutive data points, and the same model was fitted to data in each of the subsets. To identify break points in the dynamic regime, all possible combinations of zero, one, and two break points between subsets of the time series were tested, on the condition that any subset should at least have three data points to avoid overfitting. Finally, the data from the first and last subset were combined and fitted with one parameter set to determine whether the initial and final part of the time series were characterized by the same dynamic law. This scenario was tested because initial results suggested that Harmonia population dynamics had returned to previous patterns by the end of the study (see Results). Models were ranked according to Akaike s information criterion (AIC). This ranking was used to identify the number of break points that was most supported by the data, including the time of those breaks, and whether the dynamics were characterized by different values of the parameters for the first and third phase of the dynamics (Hilborn and Mangel 1997, Bolker 2008). The model-fitting was conducted with least squares using minpack.lm (Elzhov et al. 2013) in R, version (R Development Core Team 2014). Two alternative models were tested for all break combinations: the Ricker and logistic population models (Turchin 2003). Both logistic and Ricker models are single-variable discrete time models in which N(t þ 1), the population at time t þ 1, is a function of N(t), where N(t) is the population density of Harmonia observed at our site in year t. Both models contain the parameters K (the carrying capacity) and r (per capita yearly rate of increase) estimated for Harmonia during that time period. These simple models embody the second law of population dynamics, which states that the geometric population dynamics generated by the first law is necessarily constrained by density dependent feedback (Turchin 2003). They are therefore well-suited to test for changes in the fundamental parameters of density dependent population growth: the intrinsic rate of increase from one year to the next and the carrying capacity. We use discrete time models for population dynamics because our analysis focuses on the year-toyear changes in abundance. The dynamics of prey are not explicitly included in these model formulations. The discrete form of the logistic model is Nðt þ 1Þ ¼NðtÞ 1 þ r 1 NðtÞ ð1þ K and the Ricker model takes the form Nðt þ 1Þ ¼NðtÞexp r 1 NðtÞ : ð2þ K AIC was used to determine the model type and break point combination with the greatest support from the data. AIC values associated with fitting each data subset were summed for each break point combination to give a total AIC by break point combination and model type. When the total AIC values differed by more than two units, the model with the lower AIC value was considered to have a better fit. When AIC values differed by less than two units, model performance was considered to be equivalent (Burnham and Anderson 2002). The model type and break point combination with the lowest total AIC was used to estimate K and r for each time series subset. Explicit details of data subsetting and model parameterization are available in the Appendix. Plots were generated using ggplot2 (Wickham 2009). The complete model selection process and subsequent analyses are available as an R script on github (available online). 7 We tested the sensitivity of our model selection analysis procedure to variation in the sampling period by running the model selection algorithm on data that were left unculled (all sample dates) and on data culled after day 220 (8 August) instead of day 240 (28 August) and found model selection results were the same as in the original analysis, although variability increased in some approaches tested. To account for week-to-week variations in captures, we also computed the integral of the population curve for each growing season and subjected these data to an identical model selection procedure to that used for the average captures per trap described previously. This data processing methodology produced nearly identical results as the analysis using the average per trap captures and is thus not reported here, but complete details of the data preprocessing and analysis are available in the R script on github (see footnote 7). Soybean aphid population dynamics We used Cooperative Extension pest newsletters and related publications to determine soybean aphid dates of occurrence and overall yearly infestation levels in Michigan, Iowa, Illinois, and Wisconsin from 2000 to For each state, we accessed hard copy or online Cooperative Extension pest newsletter archives and checked all reports related to field crops and additional bulletins or other reports with soybean and insect in the title for June October of each year. Reports were typically available at a weekly frequency, and the 7

4 1810 CHRISTINE A. BAHLAI ET AL. Ecological Applications Vol. 25, No. 7 narrative was searched to assess the overall intensity of soybean aphid infestation in the state as a whole. Overall infestation intensity was rated based on the following scale: low, few or no aphids reported; spotty, some localized aphid populations, but action threshold (250 aphids/plant) not reached, or if reached, occurrences were noted as rare or isolated; moderate, action threshold reached in some areas of the state; and high, action threshold commonly reached across multiple areas the state. For Michigan data, we accessed hardcopies of the Michigan State University (MSU) Field Crop Advisory Team Alert Newsletter, which were provided by the MSU Integrated Pest Management (IPM) Program Coordinator. For Iowa, we accessed the online version of their Integrated Crop Management newsletter (Iowa State University Extension 2007) and soybean aphid podcasts (available online). 8 For Illinois, we accessed a bulletin published by the Illinois Extension service (University of Illinois Extension 2012), and for Wisconsin, the Integrated Pest and Crop Management newsletter (University of Wisconsin Extension 2014). We also accessed the Soybean Aphid IPM Pest Information Platform for Extension and Education (PIPE) website (available online) for auxiliary information. 9 However, these data were only used to confirm or clarify information obtained from the other sources if language used to describe soybean aphid populations was ambiguous or inconsistent. Insecticide use We obtained data on the use of four common insecticides used in soybean and labelled for soybean aphid control in the Midwestern United States for the years Using data produced by the U.S. Geological Survey, as part of the National Water- Quality Assessment program, estimates of total statewide use of cyhalothrin-lambda, esfenvalerate, imidacloprid, and thiamethoxam in soybean for each year were extracted. These data were developed using proprietary surveys of farm operations combined with USDA statistics (data available online) 10 to extrapolate for missing surveys, with methodology described by Stone (2013). Cyhalothrin-lambda and esfenvalerate are foliar-applied, pyrethroid insecticides, while imidacloprid and thiamethoxam are neonicotinoid insecticides, primarily applied to soybean seed. Because Stone (2013) did not provide data at the commodity level necessary for our analyses, pesticide use data by state and by commodity were obtained by request to the author (W. Stone, personal communication). Total area of soybean planted by year, by state, was obtained from the National Agricultural Statistics Service (see footnote 10). The estimated amounts of pesticides used in soybeans in each state and year were divided by the total area planted to soybean to estimate average pesticide use per unit land-area planted to soybean per state and year. Pesticide data were compared to extension records to determine if patterns existed between aphid infestation and pesticide use. RESULTS Harmonia population dynamics Harmonia populations were first recorded at the KBS site in 1994, and over the intervening years until 2000, populations of this species gradually rose (Fig. 1). From 2001 to 2005, we observed a dramatic two-year boom bust cycle in Harmonia abundance. In 2006, Harmonia populations remained high, but in 2007, abundance of the species dropped to very low levels. In , Harmonia populations gradually increased at the site once again, before settling into a two-year cycle in , with both a lower mean density and lower amplitude fluctuations compared to the cycling observed in Model selection analysis indicated that Ricker models consistently outperformed logistic models when applied to the same data (Table 1). The zero break point and one break point models had considerably poorer performance (i.e., higher AIC) than models with two break points, thus model selection indicated that patterns in Harmonia changed twice over the course of the study: once after 2000 and once after 2005 (Table 1, Fig. 1). Hereafter, we refer to as Phase A, as Phase B, and as Phase C. Regression coefficients r and K computed for the bestfitting, two break point Ricker model suggested an interesting pattern. In comparing Phase B to Phase A, r and K for Harmonia increased by 20% and 40%, respectively. Then, in Phase C, these coefficients returned (within standard error) to similar values as observed in Phase A (Table 1). A graphical representation of model fit (Fig. 2) further suggested very similar behavior of data during these two phases. A model selection experiment constrained to treat Phases A and C as a single data subset improved performance over the unconstrained two break point model (Table 1, dashed line in Fig. 2). Soybean aphid population dynamics Scouting and extension records indicated that after soybean aphid was first detected in the American Midwest in 2000, there were several years in which a largely synchronous, two-year boom bust cycle was observed (Table 2). In 2001, 2003, 2005, and 2007, most states reported moderate-to-high populations of this aphid, with low populations observed in 2002, 2004, and However, in Michigan, where KBS is located, high populations of soybean aphid have not been observed since 2005, and populations of this aphid have not exceeded moderate levels since 2007 (Table 2).

5 October 2015 REGIME SHIFTS IN AN INVASIVE PREDATOR 1811 FIG. 1. Population density of Harmonia axyridis (multicolored Asian lady beetle) at Kellogg Biological Station, southwestern Michigan, Break points in the time series after 2000 and after 2005 were determined by model selection as points in time where there were changes in the population dynamics of this species. Thus we divided the data into three phases: Phase A, ; Phase B, ; and Phase C, Note that phases are defined by the model parameterization used to predict the following year. For example, density in 2001 (first year of Phase B) is predicted from density in 2000, using Phase A parameters. Insecticide use Almost no insecticide applications targeting aphids were used in Midwestern soybean fields prior to 2003 (Fig. 3). Foliar pyrethroids were initially used on soybean in the Midwest during 2003, rapidly increased in use from , and then largely stabilized. Michigan led in both esfenvalerate and cyhalothrinlambda use per hectare of soybean in 2005, likely in response to very high aphid populations in this state compared to others for which we have records. Seed coated with neonicotinoid insecticide was first widely used throughout the Midwest in Thiamethoxam use per hectare of soybean rapidly increased from 2006 to 2009 in all states, although Michigan consistently had one of the highest rates of application. Imidacloprid was rarely used in soybean in the Midwest before 2009, except for Michigan in In general, increased use of foliar-applied insecticide was associated with higher levels of aphid infestation, whereas there was a trend towards lower aphid infestation in years and states in which more seed-applied insecticides were used (Fig. 4). Regression coefficients and model selection criteria for Ricker and logistic models for year-to-year population dynamics of the Asian lady beetle Harmonia axyridis at Kellogg Biological Station, Hickory Corners, Michigan, USA, TABLE 1. Model structure and breaks Subsets r K AIC Ricker (constrained) , Logistic (constrained) , Notes: Data were subset at all combinations of break points and fitted using Ricker and logistic models (Eqs. 1 and 2), where r is the yearly per capita intrinsic rate of increase and K is the carrying capacity. Constrained 2 break point models had the first and final data subsets regressed together so the two subsets would be constrained to have the same parameter values. Model performance was ranked using AIC (Akaike s information criterion). Presented in this table are the best-performing models for each model structure by break point combination. For each phase we also estimated the variance (not given).

6 1812 CHRISTINE A. BAHLAI ET AL. Ecological Applications Vol. 25, No. 7 FIG. 2. Best-fit Ricker model depicts the population density of Harmonia axyridis in three phases. The model describes population density in year t þ 1 as a function of its density in year t. Phases are as described in Fig. 1. The dashed black line represents the Ricker model structure when Phases A and C are combined into a single data set and regressed together. Points are labeled according to the year for N(t). DISCUSSION Population data for Harmonia over 20 years at the KBS support the conclusion that two shifts in the regulation of this invasive species occurred over this time period. Abrupt changes in the dynamics of this species were observed after 2000, and again after 2005 (Table 1). Perhaps most interestingly, however, is that after the second shift in Harmonia s dynamics, we observed a return to the dynamics that was observed immediately after Harmonia s initial invasion. The first identified shift in the dynamics of Harmonia can be easily explained by the arrival of soybean aphid. Previous studies have commented on the link between Harmonia s population explosion in the early 2000s with the invasion and population-cycling of soybean aphid in central North America (Bahlai and Sears 2009, Heimpel et al. 2010, Knapp et al. 2012). Soybean aphid presented a significant new resource for Harmonia in the Midwestern United States. Given a new, highly abundant, and highly suitable resource on the landscape, it is unsurprising that we observe higher estimated values of both the intrinsic rate of increase and the carrying capacity for Harmonia (Mignault et al. 2006). Not only was the landscape able to support more Harmonia, the lady beetles were potentially able to complete development more quickly and produce greater numbers of offspring in the presence of this new resource (Osawa 2000). Harmonia efficiently moves through agricultural landscapes, rapidly locating aphid colonies (Donaldson and Gratton 2007, Forbes and Gratton 2011), and it may be able to capitalize on soybean aphid as a resource to a greater extent than native lady beetle species (Bahlai et al. 2013, 2015). Given the association between Harmonia and soybean aphid, and the link between the aphid s arrival and a change in its dynamics, it seems reasonable to hypothesize that the second identified shift in Harmonia s dynamics is also linked to dynamics of soybean aphid. Our model selection results indicated that Harmonia s population regulation patterns returned to their presoybean aphid state, suggesting that the aphid no longer substantially affected the population dynamics of Harmonia. Given that the dynamic regime of Harmonia in recent years became similar to what it had been before the invasion of soybean aphid, we looked for a factor that is extrinsic to the aphid lady-beetle interaction and could have affected soybean aphid densities and then indirectly affected Harmonia. Pesticide-use records offer a plausible explanation of the observed patterns in both species. Insecticide use appears closely linked to soybean aphid s abundance. However, patterns differ between foliar- and seed- TABLE 2. Relative infestation level of soybean aphid in four Midwestern U.S. states, Year Michigan Iowa Illinois Wisconsin 2000 spotty low high moderate 2001 high moderate high high 2002 low spotty low spotty 2003 high high high high 2004 low low low low 2005 high moderate spotty moderate 2006 low low spotty spotty 2007 spotty moderate moderate high 2008 low moderate low moderate 2009 low moderate spotty moderate 2010 low spotty low low 2011 spotty low low low 2012 low low low low Notes: Infestation levels were compiled from scouting records and extension reports from each state. Aphid infestation was classified as follows: high, many fields in state surpassing economic threshold; moderate, some fields in one or more subregion exceeding economic threshold; spotty, rare fields exceeding economic threshold; and low, few or no aphids detected.

7 October 2015 REGIME SHIFTS IN AN INVASIVE PREDATOR 1813 FIG. 3. Pesticide use in soybean in four midwestern U.S. states, Data are presented as the estimated pesticide use by state divided by the total area in that state planted to soybean. Pesticide use was obtained from the U.S. Geological Survey estimated by methodology described in Stone (2013). By state, by commodity pesticide use data were obtained by request to the author. Area of soybean planted by year, by state, was obtained from the National Agricultural Statistics Service (see footnote 5). The vertical dashed line between 2005 and 2006 indicates the location of an apparent phase shift in the population dynamics of Harmonia axyridis. treatment insecticides. We observed increased usage of the foliar insecticides cyhalothrin-lambda and esfenvalerate in locations and years with higher levels of aphid infestation (Fig. 4). This trend is expected; foliar insecticides are only recommended for control of soybean aphid when aphid populations exceed 250 aphids per plant (Ragsdale et al. 2007, Tilmon et al. 2011, Hallett et al. 2014). However, we observe the opposite trend between aphid infestation and seedapplied insecticide use, indicating that the cause effect relationship may be reversed: increased usage of imidacloprid and thiamethoxam is correlated with a decline in aphid infestation (Fig. 4). This result is somewhat surprising, as several authors have remarked that seed treatments are limited in their efficacy for preventing yield loss from soybean aphids at the scale of field or plot (Ragsdale et al. 2007, Johnson et al. 2008, Seagraves and Lundgren 2012, Myers and Hill 2014). Soybean aphid s mortality due to seed-applied insecticides subsides at days after planting, decreasing to no noticeable effect to aphids colonizing soybeans later in the growing season (Seagraves and Lundgren 2012, McCarville and O Neal 2013, Myers and Hill 2014). Because economically injurious populations of soybean aphid have historically occurred later in the growing season, seed-applied insecticides do not protect the crop from late-season outbreaks (Johnson et al. 2008). However, our data show that in years and states with increased seed treatment use, outbreaks of soybean aphid were fewer and less widespread, indicating that the early impact of neonicotinoids on establishing aphid populations affected their population levels throughout the season in a similar way as described for early season predation by natural enemies (van der Werf et al. 1992,

8 1814 CHRISTINE A. BAHLAI ET AL. Ecological Applications Vol. 25, No. 7 FIG. 4. Pesticide use by soybean aphid infestation category, (see Table 2 for definitions of categories). Pesticide use was estimated by year and by U.S. state and compared to scouting records and extension records by state in a given year. Dashed lines connect the medians between each subsequent aphid infestation category, boxes represent the middle quartiles of the data, whiskers on the plots represent maximum and minimum values (excluding outliers), and dots represent outliers (observations greater than 1.5 times the upper quartile or 1.5 times less than the lower quartile, respectively). Landis and van der Werf 1997). Although these data are correlative and a variety of factors may affect aphid density, rendering this interpretation speculative, they offer a plausible explanation for the declines in soybean aphid populations in the Midwestern United States that is congruent with current knowledge on the effects of neonicotinoids on soybean aphid and of mortality during early aphid establishment on season population dynamics of aphids in field crops. In recent years in the U.S. corn belt and Great Lakes region, approximately one third of soybean acreage was planted with neonicotinoid treated seed (Myers and Hill 2014). The early-season effects of seed treatments may prevent colonization of treated soybean fields by aphids or sufficiently damp early-season population growth of aphids at the landscape scale to the point of having season-long consequences for soybean aphid abundance. An interesting implication of this observation is that area-wide suppression of soybean aphid populations may have been a direct result of the adoption of seed treatments, despite limited evidence showing the benefits of use of these treatments within individual fields. When colonizing soybean fields in the spring, a winged female soybean aphid, will settle, feed, and deposit several nymphs before taking off in search of another host patch (Ragsdale et al. 2004). This colonization behavior increases the likelihood of an individual female encountering a neonicotinoid treated field: if every third field is treated, not only is colonization less likely to be successful, it is likely that the colonizing female will herself be exposed to insecticide, limiting her chances to move on to another field. Genetic analyses suggest that spring colonization of soybean fields by soybean aphid is completed by relatively few individuals within a region (Michel et al. 2009). Thus, even though only a portion of soybean fields in a given state would have been planted with treated seed, the prevalence of these inhospitable environments may disrupt soybean aphid population cycling. A similar landscape-level inoculation effect was observed with European corn borer after the widespread adoption of Bt transgenic crops in the Midwestern United States (Hutchison et al. 2010). Our data support the assertion that the second phase shift in Harmonia s dynamics, like the decline in soybean aphid populations, could plausibly have been driven by insecticide use patterns in soybean. A rapid increase in foliar insecticide use occurred in Michigan in 2005 in response to a severe aphid outbreak in the state in that year (Fig. 3, Table 2). Two different insecticidemediated mechanisms could be involved in causing this phase shift: mortality of Harmonia as a direct result of pesticide spraying or indirect insecticide effects, due to the cascading effects of reduced availability of prey (cf. Hallmann et al. 2014). Foliar-applied pyrethroids are broad-spectrum and are associated with a negative impact on Harmonia populations when applied to soybean with high aphid populations (Ohnesorg et al. 2009, dos Santos Rodrigues et al. 2013). Harmonia is

9 October 2015 REGIME SHIFTS IN AN INVASIVE PREDATOR 1815 frequently observed foraging in soybean fields when aphid densities reach economically damaging levels (Fox et al. 2004, Rutledge et al. 2004, Hallett et al. 2014), and thus, many individuals are likely exposed to foliar insecticides applied for aphid control. However, foliar insecticide use does not fully explain the pattern we observed in the dynamics of Harmonia. If foliar insecticides had indeed impacted Harmonia s population dynamics in 2005, we would expect its numbers to crash in 2006, not 2007 as observed, and model selection would have favored a break point one year earlier. Yet, in 2006, we observed high numbers of Harmonia (Fig. 1). In fact, according to our model selection, 2005 predicted 2006 s numbers of Harmonia using parameters from Phase B with great accuracy, suggesting the factors governing its dynamics changed in the 2006 growing season. Widespread use of seed-applied insecticides, which largely started in 2006 (Thelin and Stone 2013), provides a more plausible explanation to both the decline in aphids and overall shift in the dynamics of Harmonia. Although aphids may not typically be detected in soybean fields until later in the growing season, they likely occur at low levels in soybean fields soon after soybean emerges from the soil (Welsman et al. 2007). Although seed-applied insecticides are unable to prevent yield loss from aphid outbreaks for individual fields later in the growing season (Johnson et al. 2008, Johnson et al. 2009), the early-season activity of the seed treatments significantly limit the growth rate of soybean aphids resulting in lowered season-long exposure to aphids (Johnson et al. 2009, McCarville and O Neal 2013). We have shown that seed-applied insecticides are correlated with lower numbers of soybean aphid at a landscape or regional scale (Fig. 4). By preventing the establishment and reducing the growth rate of soybean aphid early in the growing season, regions with substantial use of seedapplied insecticides would experience less field-to-field migration by soybean aphid later in the growing season and fewer late-season aphid outbreaks, providing fewer resources for and contributing to changes in dynamics of aphid predators like Harmonia. Harmonia typically undergoes two to three generations per year, but when prey are scarce, lays fewer eggs and has increased larval mortality (Koch 2003). A lack of early-season prey availability and an associated decrease in late-season outbreaks would limit Harmonia s ability to reproduce in soybean fields early in the growing season and account for its population regulation shift starting in It is possible that other factors could have caused a similar pattern to the one we observed. These factors could include the arrival of an additional aphid predator to the system. In 2007, an exotic parasitoid wasp, Aphelinus certus, was observed parasitizing soybean aphid at a high rate in Ontario, Canada, geographically adjacent to Michigan (Frewin et al. 2010). However, given that A. certus was detected through much of the North American range of soybean aphid as early as 2005, it is likely that the parasitoid arrived with soybean aphid or soon after (Desneux et al. 2009, Heimpel et al. 2010). Thus, if the decline of soybean aphid and shift in Harmonia s dynamics were primarily driven by the arrival of this parasitoid, we would have expected to see shifts in dynamics prior to 2005, or a slow shift in dynamics, rather than discrete break points. Furthermore, surveys of the natural enemy community in soybean fields within the Midwest have revealed limited evidence of parasitism by both native and exotic parasitoids (Schmidt et al. 2008, Gardiner et al. 2009). This analysis suggests that the widespread use of neonicotinoid seed treatments has contributed to the reduction of both a herbivorous crop pest and its primary predator, an invasive lady beetle species. Neonicotinoid insecticides have been implicated with the decline of insectivorous birds in The Netherlands via a trophic mechanism similar to that which we observe for Harmonia s decline (Hallmann et al. 2014). However, although bird decline is largely considered a uniformly negative environmental consequence, a decline in abundance of an invasive predator like Harmonia will meet with a more mixed valuation of ecological consequences (i.e., by loss of both biocontrol and nuisance effects, and potential relaxation of its effects on native coccinellid species and ecological food webs). Given this and other recent controversies about nontarget effects in the use of insecticides, particularly neonicotinoid seed treatments on bees (e.g., Cresswell 2011, Blacquière et al. 2012, Whitehorn et al. 2012), perceptions about efficacy towards soybean pests (Myers and Hill 2014) and potential changes to the regulation of this class of insecticides, alternative methods for achieving similar suppressive effects warrant investigation. Landscape-level suppression of soybean aphid and Harmonia may be possible, or even improved upon, by using varieties of soybeans that are aphid-resistant. Soon after the discovery of the soybean aphid in North America, aphid resistance (resistance to Aphis glycines, or Rag genes) was discovered in soybean germplasm. Although use of a single gene (i.e., Rag1 or Rag2) has been met with limited success in controlling soybean aphid populations (Hesler et al. 2013), incorporation of two genes within a single soybean line prevented aphid populations from reaching action thresholds (Wiarda et al. 2012). These resistant soybean lines have the capacity to produce season-long suppression of soybean aphid populations while limiting nontarget effects (Lundgren et al. 2009, McCarville and O Neal 2013, McCarville et al. 2014). Furthermore, the management of soybean aphids is not improved when a seed treatment is applied to varieties incorporating multiple resistance genes. As such, it is possible that the adoption of multigene-resistant soybean varieties may allow neonicotinoid seed treatments to be reduced

10 1816 CHRISTINE A. BAHLAI ET AL. Ecological Applications Vol. 25, No. 7 or eliminated, while maintaining low populations of both soybean aphid and Harmonia. CONCLUSIONS We have shown that the invasion of a herbivore (soybean aphid) and the consequent management of this herbivore by human intervention using insecticides, several years later, resulted in dramatic and abrupt changes in the year to year dynamics of a predator (Harmonia) that can efficiently exploit the aphid food resource. The changes in the dynamics of the predator were not smooth, but can be described as changes in dynamic regime, even though the changes in aphid density and management were in principle gradual. Currently, the dynamics of the exotic lady beetle Harmonia are no longer being driven by soybean aphid in the North American Midwest. In fact, our analysis suggests that soybean aphid is no longer a significant resource for Harmonia on the landscape. Insecticides, specifically the prophylactic use of neonicotinoids applied to soybean seed, may be important drivers in this process and demonstrate that regime changes driven by invasion processes and occurring at higher trophic levels may, in fact, be reversible by management. If invasive pests are well controlled, their effects at other trophic levels can be minimized, as shown by the fact that dynamics of Harmonia returned to its initial pattern. ACKNOWLEDGMENTS The authors gratefully acknowledge J. Megan Woltz and Julia Perrone for their assistance with data collection, Wesley Stone for sharing and aiding in the interpretation of the United States Geological Survey National Water-Quality Assessment data, and the helpful comments of two anonymous reviewers. This work was supported by the National Science Foundation Long Term Ecological Research Program at the Kellogg Biological Station (NSF DEB ) and funded in part with support from the U.S. Department of Energy s Great Lakes Bioenergy Research Center (DOE BER Office of Science DE-FC02-07ER64494) and by the USDA National Institute of Food and Agriculture and Michigan State University AgBio- Research. LITERATURE CITED Bahlai, C. A., M. Colunga-Garcia, S. H. Gage, and D. A. Landis Long term functional dynamics of an aphidophagous coccinellid community are unchanged in response to repeated invasion. PLoS ONE 8:e Bahlai, C. A., M. Colunga-Garcia, S. H. Gage, and D. A. Landis The role of exotic species in the decline of native ladybeetle populations: evidence from long-term monitoring. Biological Invasions 17: Bahlai, C. A., and M. K. Sears Population dynamics of Harmonia axyridis and Aphis glycines in Niagara Peninsula soybean fields and vineyards. Journal of the Entomological Society of Ontario 140: Biggs, R., S. R. Carpenter, and W. A. Brock Turning back from the brink: detecting an impending regime shift in time to avert it. Proceedings of the National Academy of Sciences USA 106: Blacquie` re, T., G. Smagghe, C. M. van Gestel, and V. Mommaerts Neonicotinoids in bees: a review on concentrations, side-effects and risk assessment. Ecotoxicology 21: Bolker, B. M Ecological models and data in R. Princeton University Press, Princeton, New Jersey, USA. Brown, M. W., and S. S. Miller Coccinellidae (Coleoptera) in apple orchards of eastern West Virginia and the impact of invasion by Harmonia axyridis. Entomological News 109: Burnham, K. P., and D. R. Anderson Model selection and multimodel inference: a practical information-theoretic approach. Second edition. Springer ScienceþBusiness Media, New York, New York, USA. Carpenter, S. R., and W. A. Brock Rising variance: a leading indicator of ecological transition. Ecology Letters 9: Carpenter, S. R., W. A. Brock, J. J. Cole, J. F. Kitchell, and M. L. Pace Leading indicators of trophic cascades. Ecology Letters 11: Colunga-Garcia, M., and S. H. Gage Arrival, establishment, and habitat use of the multicolored Asian lady beetle (Coleoptera: Coccinellidae) in a Michigan landscape. Environmental Entomology 27: Colunga-Garcia, M., S. H. Gage, and D. A. Landis Response of an assemblage of Coccinellidae (Coleoptera) to a diverse agricultural landscape. Environmental Entomology 26: Costamagna, A. C., and D. A. Landis Predators exert top-down control of soybean aphid across a gradient of agricultural management systems. Ecological Applications 16: Costamagna, A. C., and D. A. Landis Quantifying predation on soybean aphid through direct field observations. Biological Control 42: Cresswell, J A meta-analysis of experiments testing the effects of a neonicotinoid insecticide (imidacloprid) on honey bees. Ecotoxicology 20: Crowder, D., and W. Snyder Eating their way to the top? Mechanisms underlying the success of invasive insect generalist predators. Biological Invasions 12: Desneux, N., R. Barta, K. Hoelmer, K. Hopper, and G. Heimpel Multifaceted determinants of host specificity in an aphid parasitoid. Oecologia 160: DiFonzo, C., and R. Hines Soybean aphid in Michigan. Extension Bulletin E Michigan State University, East Lansing, Michigan, USA. Donaldson, J. R., and C. Gratton Antagonistic effects of soybean viruses on soybean aphid performance. Environmental Entomology 36: dos Santos Rodrigues, A. R., A. F. Spíndola, J. E. de Morais Oliveira, and J. B. Torres Dietary effects upon biological performance and lambda-cyhalothrin susceptibility in the multicolored Asian lady beetle, Harmonia axyridis. Phytoparasitica 41: Elzhov, T. V., K. M. Mullen, A.-N. Spiess, and B. Bolker minpack.lm: R interface to the Levenberg-Marquardt Nonlinear least-squares algorithm found in minpack R package, version minipack.lm/index.html Folke, C., S. Carpenter, B. Walker, M. Scheffer, T. Elmqvist, L. Gunderson, and C. Holling Regime shifts, resilience, and biodiversity in ecosystem management. Annual Review of Ecology, Evolution, and Systematics 35: Forbes, K. J., and C. Gratton Stable isotopes reveal different patterns of inter-crop dispersal in two ladybeetle species. Ecological Entomology 36: Fox, T. B., D. A. Landis, F. F. Cardoso, and C. D. Difonzo Predators suppress Aphis glycines Matsumura population growth in soybean. Environmental Entomology 33: Frewin, A. J., Y. Xue, J. A. Welsman, B. A. Broadbent, A. W. Schaafsma, and R. H. Hallett Development and

11 October 2015 REGIME SHIFTS IN AN INVASIVE PREDATOR 1817 parasitism by Aphelinus certus (Hymenoptera: Aphelinidae), a parasitoid of Aphis glycines (Hemiptera: Aphididae). Environmental Entomology 39: Galvan, T. L., E. C. Burkness, Z. Vickers, P. Stenberg, A. K. Mansfield, and W. D. Hutchison Sensory-based action threshold for multicolored Asian lady beetle-related taint in winegrapes. American Journal of Enology and Viticulture 58: Gardiner, M. M., D. A. Landis, C. Gratton, C. D. DiFonzo, M. O Neal, J. M. Chacon, M. T. Wayo, N. P. Schmidt, E. E. Mueller, and G. E. Heimpel Landscape diversity enhances biological control of an introduced crop pest in the north-central USA. Ecological Applications 19: Hallett, R. H., C. A. Bahlai, Y. Xue, and A. W. Schaafsma Incorporating natural enemy units into a dynamic action threshold for the soybean aphid, Aphis glycines (Hemiptera: Aphididae). Pest Management Science 70: Hallmann, C. A., R. P. B. Foppen, C. A. M. van Turnhout, H. de Kroon, and E. Jongejans Declines in insectivorous birds are associated with high neonicotinoid concentrations. Nature. Harmon, J. P., E. Stephens, and J. Losey The decline of native coccinellids (Coleoptera: Coccinellidae) in the United States and Canada. Journal of Insect Conservation 11: Heimpel, G., L. Frelich, D. Landis, K. Hopper, K. Hoelmer, Z. Sezen, M. Asplen, and K. Wu European buckthorn and Asian soybean aphid as components of an extensive invasional meltdown in North America. Biological Invasions 12: Hesler, L. S., et al Performance and prospects of Rag genes for management of soybean aphid. Entomologia Experimentalis et Applicata 147: Hilborn, R., and M. Mangel The ecological detective: confronting models with data. Princeton University Press, Princeton, New Jersey, USA. Hulme, P. E Trade, transport and trouble: managing invasive species pathways in an era of globalization. Journal of Applied Ecology 46: Hutchison, W. D., et al Areawide suppression of European corn borer with Bt maize reaps savings to non-bt maize growers. Science 330: Iowa State University Extension Integrated crop management news. taxonomy/term/578 Ives, A. R., and S. R. Carpenter Stability and diversity of ecosystems. Science 317: Johnson, K. D., M. E. Neal, J. D. Bradshaw, and M. E. Rice Is preventative, concurrent management of the soybean aphid (Hemiptera: Aphididae) and bean leaf beetle (Coleoptera: Chrysomelidae) possible? Journal of Economic Entomology 101: Johnson, K. D., M. E. O Neal, D. W. Ragsdale, C. D. Difonzo, S. M. Swinton, P. M. Dixon, B. D. Potter, E. W. Hodgson, and A. C. Costamagna Probability of cost-effective management of soybean aphid (Hemiptera: Aphididae) in North America. Journal of Economic Entomology 102: Kenis, M., M.-A. Auger-Rozenberg, A. Roques, L. Timms, C. Pe re, M. W. Cock, J. Settele, S. Augustin, and C. Lopez- Vaamonde Ecological effects of invasive alien insects. Biological Invasions 11: Knapp, A. K., et al Past, present, and future roles of long-term experiments in the LTER Network. Bioscience 62: Koch, R., and T. Galvan Bad side of a good beetle: the North American experience with Harmonia axyridis. Biocontrol 53: Koch, R. L The multicolored Asian lady beetle, Harmonia axyridis: a review of its biology, uses in biological control, and non-target impacts. Journal of Insect Science 3:1 16. Koch, R. L., R. C. Venette, and W. D. Hutchison Invasions by Harmonia axyridis (Pallas)(Coleoptera: Coccinellidae) in the western hemisphere: implications for South America. Neotropical Entomology 35: Landis, D., and W. van der Werf Early-season predation impacts the establishment of aphids and spread of beet yellows virus in sugar beet. Entomophaga 42: Landis, D. A., and S. H. Gage Arthropod diversity and pest suppression in agricultural landscapes. Pages in S. K. Hamilton, J. E. Doll, and G. P. Robertson, editors. The ecology of agricultural landscapes: long-term research on the path to sustainability, Oxford University Press, New York, New York, USA. Lundgren, J., L. Hesler, K. Tilmon, K. Dashiell, and R. Scott Direct effects of soybean varietal selection and Aphis glycines-resistant soybeans on natural enemies. Arthropod- Plant Interactions 3:9 16. Mack, R. N., D. Simberloff, W. Mark Lonsdale, H. Evans, M. Clout, and F. A. Bazzaz Biotic invasions: causes, epidemiology, global consequences, and control. Ecological Applications 10: McCarville, M. T., and M. E. O Neal Soybean aphid (Aphididae: Hemiptera) population growth as affected by host plant resistance and an insecticidal seed treatment. Journal of Economic Entomology 106: McCarville, M. T., M. E. O Neal, B. D. Potter, K. J. Tilmon, E. M. Cullen, B. P. McCornack, J. F. Tooker, and D. A. Prischmann-Voldseth One gene versus two: a regional study on the efficacy of single gene versus pyramided resistance for soybean aphid management. Journal of Economic Entomology 107: McClure, M Potential of the Asian predator, Harmonia axyridis Pallas (Coleoptera: Coccinellidae), to control Matsucoccus resinosae Bean and Godwin (Homoptera: Margarodidae) in the United States. Environmental Entomology 16: Michel, A. P., W. Zhang, J. Kyo Jung, S.-T. Kang, and M. A. Rouf Mian Population genetic structure of Aphis glycines. Environmental Entomology 38: Mignault, M.-P., M. Roy, and J. Brodeur Soybean aphid predators in Québec and the suitability of Aphis glycines as prey for three Coccinellidae. Biocontrol 51: Myers, C., and E. Hill Benefits of neonicotinoid seed treatments to soybean production. United States Environmental Protection Agency, Washington, D.C., USA. www2.epa.gov/pollinator-protection/benefits-neonicotinoidseed-treatments-soybean-production Nakazawa, T., S. M. Satinover, L. Naccara, L. Goddard, B. P. Dragulev, E. Peters, and T. A. E. Platts-Mills Asian ladybugs (Harmonia axyridis): a new seasonal indoor allergen. Journal of Allergy and Clinical Immunology 119: Ohnesorg, W. J., K. D. Johnson, and M. E. O Neal Impact of reduced-risk insecticides on soybean aphid and associated natural enemies. Journal of Economic Entomology 102: Osawa, N Population field studies on the aphidophagous ladybird beetle Harmonia axyridis (Coleoptera: Coccinellidae): resource tracking and population characteristics. Population Ecology 42: Persson, L., P.-A. Amundsen, A. M. De Roos, A. Klemetsen, R. Knudsen, and R. Primicerio Culling prey promotes predator recovery: alternative states in a whole-lake experiment. Science 316: Pervez, A., and Omkar Ecology and biological control applications of multicoloured Asian ladybird, Harmonia axyridis: a review. Biocontrol Science and Technology 16: