Description of study site and classification of habitats

All studies occurred at Michigan State University’s Kellogg Biological Station (KBS) in southwestern Michigan (42°24’N, 85°24’W) at 288 m elevation. 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 forest sites. See http://lter.kbs.msu.edu/research/long-term-experiments/main-cropping-system-experiment/ and http://lter.kbs.msu.edu/research/long-term-experiments/successional-and-forest-sites/ for a detailed experimental design. The MCSE consists of seven treatments including treatments of annual field crops (maize, soybeans and wheat) in annual rotation under 4 levels of management intensity (conventional, no-till, reduced input, and biologically based), as well as alfalfa, poplar and early successional vegetation (i.e. abandoned agricultural fields). Each treatment is replicated six times with individual plot sizes of 1 ha. Observations are taken from five permanent sampling stations within each treatment-replicate combination. Instead of grouping by management intensity for this analysis, coccinellid abundance data from the MCSE were grouped by the particular host crop planted in that treatment-replicate combination in a given observation-year combination. Forest sampling began in 1993, at sites within 3 km of the MCSE on the KBS. Forest types included: 40-60 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 are monitored at five sampling stations, as in the MCSE. The LTER experiment continues to date. Detailed agronomic records for the site are available at http://lter.kbs.msu.edu/datatables/150. Coccinellid abundance data is available online at http://lter.kbs.msu.edu/datatables/67. A relational database in MS Access® integrating coccinellid data with the relevant environmental and agronomic records used in this analysis is available from the authors by request.

To monitor coccinellids, un-baited two-sided yellow cardboard sticky cards (Pherocon, Zoecon, Palo Alto CA) are suspended at 1.2 m above ground at each sampling station [29]. Each week, coccinellids on the sticky cards are identified using a pictorial key, counted, recorded, and then removed from sticky cards. Sticky cards are replaced every two weeks. The length of the sampling period within the growing season averaged 14 weeks (with a standard error of one week, and minimum and a maximum of 8 and 22 weeks respectively). Variation depended on the availability of labor, crop type, and weather.

For analyses and data presentation, individual habitats (i.e. dominant plant community occurring in a treatment block) were grouped into three habitat types: annual crops, perennial crops, and forest habitats. Annual crops were all dominated by annual herbaceous plants, grown in near monoculture, and included maize, wheat, and soybeans. Habitats classed as perennial crops included poplar, alfalfa, and managed early successional. Managed early successional habitats were included in this group because, although this community is not a crop per se, it is a community resulting from abandoned crop land, occurring within a crop matrix, and managed to maintain their early successional stage (e.g., yearly spring burnings). Finally, forest habitats included coniferous, deciduous, and successional forest.

Coccinellid abundance data

The relative abundance of 13 species of aphidophagous coccinellids (Table 1) as measured by sticky trap capture of adults was used in the analysis. Species were included if they were recorded in the surveys, and had been recorded in the literature as consuming aphids. The majority of coccinellids recorded are primarily aphidophagous, but Coleomegilla maculata, an omnivorous species that also feeds on pollen [30], and Chilocorus stigma, a scale specialist that incidentally feeds on aphids [31], were included in the analysis. Two other coccinellids, Brachiacantha ursina, and Psyllobora vigintimaculata, are only known to eat scale insects and fungi, respectively [32], and were excluded from the analysis.

Data were examined at two different temporal resolutions to gain insight into both real-time interactions and between season trends. First, the sum of all captures from each sampling station in a given week, in a given treatment by replicate combination, was called ‘weekly average,’ and the sum of all captures in a given treatment over a calendar year, was referred to as ‘yearly total.’ Data were further subdivided into sets of ‘all coccinellids’ and ‘native species only’ to detect differences in diversity and function of native coccinellids compared to the community as a whole. To aide in data visualization, a Gaussian smoother with a span of 3 units was used to produce smoothing curves. To reduce the likelihood of erroneous conclusions from a large dataset with numerous sources of variation, α=0.01 was observed for all analyses.

Diversity analysis

The Shannon Index (Shannon’s H) was used to examine trends in diversity of the coccinellid community at the site over time [33]. A Shannon Index was computed for the species distribution at each observation and then this value was used as the dependent variable in several analyses examining the Shannon Index by each habitat type. One way ANOVA or, if data did not meet assumptions of ANOVA, Kruskal-Wallis one way analysis of variance on ranks were performed using SigmaPlot 11.0.0.75 For Windows (Systat Software Inc.) for Shannon Index by habitat for both datasets. Weighted least square regressions were conducted in R 2.15.3 [34] to detect linear trends in Shannon Indices over the observation period. Regressed data points were weighted by number of traps reporting in a given replicate-week combination, or yearly total number of traps in a habitat, for weekly and yearly datasets, respectively.

Herbivore suppression potential model

The herbivore suppression potential of the observed community was modeled by weighting each species by its relative voracity, after Bahlai et al. [17]. Herbivore suppression potential is computed in natural enemy units (NEU), where 1 NEU is the number of predators or parasitoids required to kill 100 of a particular pest insect in 24h. Thus,

for N species, where n_i is number of individuals of species i, and V_i is the voracity of species i, expressed as the number of target prey (in this case, aphids) consumed in 24h, divided by 100.

Voracity of adult coccinellids

To estimate the voracity of each coccinellid species, literature values of average aphid consumption rates were used. If a study reported aphid consumption rates for each sex for a given species, a 1:1 sex ratio was assumed and the average of the consumption rates are reported. Where multiple sources were available, the highest estimate of voracity was used, except when the highest estimate was dramatically higher than all other published values; then the top two estimates were averaged. The voracity values used to estimate herbivore suppression potential for all aphidophagous coccinellids examined in this study (obtained from the literature) are provided in Table 2.

Analysis of herbivore suppression services and relationship with biodiversity

To detect linear trends in herbivore suppression potential over the observation period, one way ANOVA or, if data did not meet assumptions of ANOVA, Kruskal-Wallis one way analysis of variance on ranks were performed for herbivore suppression potential by habitat for both datasets. Least square regressions weighted by the number of reporting traps were performed herbivore suppression potential were examined using data at the weekly resolution only. A least-square linear regression analysis was performed comparing the Shannon Index to herbivore suppression potential for captured coccinellids for each habitat type on the weekly average dataset. Year was treated as a factor in the analysis to account for year-to-year variability in overall coccinellid occurrence. ANOVA was performed on the regression model to determine if the interaction of Shannon diversity and herbivore suppression potential differed between habitats.

Coccinellid abundance data

Between 1989 and 2012, 739,047 individual observations were recorded (species by trap) in the KBS LTER coccinellid database, 61,377 traps were monitored, and 57,813 aphidophagous coccinellids corresponding to thirteen species were captured.

Community structure and dynamics

Coccinellid community structure varied markedly over the course of the study (Figure 1). The proportion of total captured coccinellids for each species from 1989 to 2012 is given in Table 1, and the proportional abundance by year of species accounting for ≥1% of total captures is given in Figure 1. Over the entire study, C. septempunctata, and more recently, H. axyridis comprised the two dominant species, together accounting for more than 69% of total captures. Prior to 1996, C. maculata was frequently observed to be co-dominant with C. septempunctata, but has declined in importance since, typically only representing greater than 20% of the captures in years when maize is planted at the site. The relative abundance of each of these three dominant species varied from year to year (Figure 1). The two most recent additions to the community, H. variegata and P. quatuordecimpunctata, are rapidly increasing in dominance. Although absolute proportions of each species captured varied dramatically by year, the community was increasingly dominated by exotic species over the course of the study. Native species represented 49.8% of captures in the first five years of the study. However, between 2008 and 2012, native species represented only 27.9% of the captures, with more than two thirds of these captures corresponding to C. maculata.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Dominant coccinellid species varies from year to year.

Proportional abundance of most common coccinellid species by year. Data were collected at Kellogg Biological Station, Hickory Corners, MI, 1989-2012.

https://doi.org/10.1371/journal.pone.0083407.g001

Total coccinellid abundance also varied widely by year over the course of the study (Figure 2). On average, coccinellid captures were highest in annual crops, followed closely by captures in perennial crops, while the trapping rate in forest habitats was consistently lower than that in either annual or perennial crop habitats (Figure 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Year-to-year variation in overall coccinellid captures.

Average trap captures per year of aphidophagous coccinellids at Kellogg Biological Station, Hickory Corners, MI, 1989-2012, in the annual and perennial crop habitats and in the adjacent forest habitat.

https://doi.org/10.1371/journal.pone.0083407.g002

Diversity analysis

Shannon diversity varied between habitat types at the weekly resolution, but not at the yearly resolution of the data (weekly: H₂ =860.7, p<0.01, yearly: H₂=7.03, p=0.03, Table 3). In a given week, the greatest diversity was observed in annual crops and the lowest diversity was observed in forest habitats. A positive linear trend in Shannon diversity by year was observed at the weekly resolution in annual and perennial crops (annual: slope=0.0102 ±0.0008, p<0.01; perennial: slope=0.0053± 0.0008, p<0.01), but no trends were observed in forest (Figure 3A), nor in Shannon diversity over time within individual habitat groups at the yearly resolution (Figure 3B). When Shannon diversity of only native species was examined at the weekly resolution, a significant linear decrease over time was observed in annual and perennial crop habitats (annual: slope=-0.0031 ±0.0004, p<0.01; perennial: slope=-0.0058± 0.0004, p<0.01), however, no linear relationship was observed in forest habitats (Figure 3C). At a yearly resolution, no significant trends were observed in native coccinellid Shannon diversity (Figure 3D).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Trends in diversity in the coccinellid community.

Shannon index of aphidophagous coccinellid community by habitat type at Kellogg Biological Station, Hickory Corners, MI, 1989-2012, showing observed values (symbols) and 3 yr smoothed (lines) data. A) Shannon index analyzed at weekly resolution, all species. B) Shannon index analyzed at yearly resolution, all species. C) Shannon index analyzed at weekly resolution, native species only. D) Shannon index analyzed at yearly resolution, native species.

https://doi.org/10.1371/journal.pone.0083407.g003

Herbivore suppression services and relationship with biodiversity

Herbivore suppression potential of the full coccinellid community varied between habitats (Kruskal-Wallis, H₂=1610.6, p<0.01, Table 3). Herbivore suppression potential was highest in annual crops, and lowest in forests.

No significant trend was observed in the modeled herbivore suppression potential of the coccinellid community over time in any of the habitat groups (Figure 4A). Herbivore suppression potential due to native species exhibited a significant linear decrease over time in annual and perennial crop habitats (annual: slope=-0.0030 ±0.0007, p<0.01; perennial: slope=-0.0074± 0.0005, p<0.01) but no trend was observed in forest habitats (Figure 4B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Trends in herbivore suppression potential of the coccinellid community.

Herbivore suppression potential (in natural enemy units, NEU) of aphidophagous coccinellid community by habitat type at Kellogg Biological Station, Hickory Corners, MI, 1989-2012, showing observed values (symbols) and 3 yr smoothed (lines) data. A) Herbivore suppression potential of all species. B) Herbivore suppression potential due to native species only. Note that the scale of the chart excludes one outlier: a value of 0.95 NEU in annual crops in 1999.

https://doi.org/10.1371/journal.pone.0083407.g004

Herbivore suppression potential was examined as a function of Shannon diversity index in three habitat types (Figure 5). The relationship between herbivore suppression potential and Shannon diversity varied between habitats (F _{2, 9374} = 24.2, p<0.01); the relationship was positive in perennial and forest habitats (slope =0.24±0.4, 0.47±0.9, respectively) but no significant relationship was found in annual habitats.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Herbivore potential varies with diversity, except in annual crops.

Herbivore suppression potential as a function of Shannon diversity index in three habitat types. For clarity, plotted points represent the average herbivore suppression potential per trap per year vs. the average Shannon index for each repetition by a crop type by sampling week combination in a given year. Regression lines are based on weighted regression of weekly observations. Regression lines marked with an asterisk indicate significant slopes at α=0.01.

https://doi.org/10.1371/journal.pone.0083407.g005

Coccinellid community dynamics

The pattern of long-term co-dominance between several species observed in this study is common in coccinellid communities (i.e., most are dominated by two to four species) although the composition of the group of dominant species may vary with location and time [35]. This co-dominance pattern is not dependent on the invasion or disturbance status of the guild. Similar co-dominance patterns have been observed in a diversity of pre-invasion coccinellid communities [19,22,24,36,37], as well as post-invasion communities [13,25].

Certain coccinellid species tend to dominate captures in areas where they are established. In our study, for instance, C. septempunctata, H. axyridis and C. maculata were among the dominant species in the coccinellid communities when they were present in the community [2,14,23,36,38,39]. P. quatuordecimpunctata has been increasing in its relative abundance since its arrival in the system, and it may join the aforementioned co-dominant species, as it, too, is known to co-dominate the coccinellid community in its native range [40].

Biodiversity and herbivore suppression potential

Net biodiversity of the coccinellid community changed very little over the course of our study. In annual and perennial crop habitats, we observed a small increase in weekly, but not yearly Shannon diversity despite the introduction of three exotic species over the course of the study (Figure 3A, B). Diversity of natives declined in annual and perennial crop habitats at a weekly resolution, and no net change to native diversity was observed over the course of this study at the yearly resolution. This suggests that although native species are still present and just as likely to be captured over the course of a year, they are less likely to be captured within a given week. This result is similar to what has been observed in previous studies. In a review by Harmon et al. [2] combining the observations of several North American coccinellid invasions, the number of native species in an assemblage did not change in response to the addition of an exotic species (i.e., it was rare to observe a complete elimination of a native species associated with the introduction of an exotic species). The diversity index associated with native species only decreased slightly, on average, after the establishment of an exotic species [2].

Modeled herbivore suppression potential of the community also changed very little in response to invasion (Figure 4). Acorn [41] suggested that the interpretation of coccinellid invasion data has been influenced by attitudes against ‘biological pollution’ even though ecological change (of the kind resulting from native species being displaced by exotics) does not yield a loss of function. Our data support this assertion, at least from the standpoint of preserving the function of aphidophagy. We estimated very little change in the long term herbivore suppression potential across our experimental landscape over the course of the study (Figure 4A). However, there was a period of increased variability in herbivore suppression potential immediately after the introduction of soybean aphid, Aphis glycines, in 2000. This introduction is known to change the resource structure for coccinellids at landscape scales [42]. During the acute invasion phase of this aphid from 2000 to 2006, coccinellid densities reached higher peaks, and were even more variable than what is typically observed (Figure 2). At this time, a brief depression in overall diversity of the coccinellid community was also observed (Figure 3B), as a single exotic species (H. axyridis) became hyper-dominant. Soybean aphid is a preferred prey item of H. axyridis and the arrival of this aphid biotically facilitated the invasion of this exotic coccinellid [43]. Soybean aphid outbreaks have since decreased in frequency, and there is evidence that, since 2007, this prey item is no longer acting as a significant driver of population dynamics of H. axyridis [44].

Diversity was significantly and positively related to herbivore suppression potential in perennial and forest habitats, but was uncorrelated in annual crops (Figure 5). This observation has important implications for biological control in these systems. When coccinellid species are partitioned into spatiotemporal niches, intraspecific competition will be more important than interspecific competition in the population dynamics of a species, intraguild predation will be uncommon, and prey suppression will increase with diversity of the community [7]. The exception to this may be in landscapes with less plant diversity, such as annual cropping systems, which provide a lower diversity of habitats and, in turn, prey species. In this case, a coccinellid species particularly adapted to the particular prey resource available may essentially dominate the entire habitat, to the exclusion of other species. However, functional diversity, that is, the partitioning of resources between consumers with different resource use patterns, is more important than simple biodiversity in determining the level of resource exploitation when resources themselves are diverse [45].

Sampling methodology

Fewer coccinellids were collected in forest habitats than in perennial or annual habitats over the duration of the study. This might be due, at least in part, to decreased efficiency of our trapping design when located in arboreal habitats (see methods). Nevertheless, more coccinellids were captured in managed early successional habitats (within the Main Cropping System Experiment) than in successional forests. The fact that these two types of successional habitats were structurally similar may indicate that indeed fewer coccinellids were present in forest habitats. Although yellow sticky cards are imperfect estimators of relative abundance of coccinellids [46], they represent a cost effective, consistent sampling unit. This characteristic makes them a useful methodology when employed in long term studies, which may be otherwise biased by personnel turnover or inconsistent sampling. Also, the fact they are permanently placed in the field may increase the probability of capturing rare species as they move through the landscape. It is important to note that between habitat and between species comparisons may be affected by the varying trap efficiency for a given species in a given habitat, and thus the strongest evidence for ecological change results from within species, within habitat trends.