https://orcid.org/0000-0002-8741-5617
Figures
Abstract
Plot size is of practical importance in any integrated pest management (IPM) study that has a field component. Such studies need to be conducted at a scale relevant to species dynamics because their abundance and distribution in plots might vary according to plot size. An adequate plot size is especially important for researchers, technology providers and regulatory agencies in understanding effects of various insect control technologies on non-target arthropods. Plots that are too small might fail to detect potential harmful effects of these technologies due to arthropod movement and redistribution among plots, or from untreated areas and outside sources. The Arizona cotton system is heavily dependent on technologies for arthropod control, thus we conducted a 2-year replicated field experiment to estimate the optimal plot size for non-target arthropod studies in our system. Experimental treatments consisted of three square plot sizes and three insecticides in a full factorial. We established three plot sizes that measured 144 m2, 324 m2 and 576 m2. For insecticide treatments, we established an untreated check, a positive control insecticide with known negative effects on the arthropod community and a selective insecticide. We investigated how plot size impacts the estimation of treatment effects relative to community structure (27 taxa), community diversity, individual abundance, effect sizes, biological control function of arthropod taxa with a wide range of mobility, including Collops spp., Orius tristicolor, Geocoris spp., Misumenops celer, Drapetis nr. divergens and Chrysoperla carnea s.l.. Square 144 m2 plots supported similar results for all parameters compared with larger plots, and are thus sufficiently large to measure insecticidal effects on non-target arthropods in cotton.
Our results are applicable to cotton systems with related pests, predators or other fauna with similar dispersal characteristics. Moreover, these results also might be generalizable to other crop systems with similar fauna.
Citation: Bordini I, Naranjo SE, Fournier A, Ellsworth PC (2023) Spatial scale of non-target effects of cotton insecticides. PLoS ONE 18(5): e0272831. https://doi.org/10.1371/journal.pone.0272831
Editor: Sean Michael Prager, University of Saskatchewan College of Agriculture and Bioresources, CANADA
Received: July 26, 2022; Accepted: March 5, 2023; Published: May 10, 2023
This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Data Availability: DOI: https://doi.org/10.25422/azu.data.21948254 The data has been deposited at the University of Arizona Research Data Depository.
Funding: This work was supported by Western Integrated Pest Management Center (IB, PE, SN, AF), Western Sustainable Agriculture Research and Extension (IB, PE, SN, AF), Arizona Cotton Growers Association (PE), United States Department of Agriculture – National Institute of Food and Agriculture (Extension Implementation Program, award number 2017-70006-27145) (PE, AF), Cotton Incorporated (PE) and United States Department of Agriculture, Agricultural Research Service- Arid Land Agricultural Research Center (SN). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Plot size is of practical importance for anyone designing experiments involving mobile species, regardless of the study purpose. Because natural arthropod communities range freely over large agricultural areas, observations from small areas could be quite different from the dynamics and patterns that occur in large areas. Thus, plots need to be as large as practically possible for experiments to be applicable to commercial scales [1–3].
Choosing an adequate plot size is particularly important in non-target arthropod assessments involving insect control technologies, e.g. insecticides and transgenic Bt crops. Several authors have demonstrated that the distribution and abundance of arthropods in plots varies according to plot size [1–6]. This variation can be especially problematic if plots are too small to detect harmful effects of these technologies on non-target arthropods due to their movement and redistribution among plots, or from untreated areas and outside sources [2].
However, larger plots necessarily result in higher costs associated with land rental, water, labor and equipment, and might not be tenable when there is a limited amount of seed in the case of regulated trials. Plots need to be as large as practically possible while balancing the goals of accurately measuring gross effects on non-target arthropods and using on an area suitable for the ecological attributes of the arthropod community, especially mobility [2, 7].
Some authors have suggested that few studies have been conducted in field plots of sufficient size to understand potential disruptions of insecticidal technologies on biological control agents and other non-target organisms [3, 8–10]. Such studies need to be conducted at a scale that is relevant to species dynamics in any integrated pest management (IPM) study that has a field component, and to allow regulatory bodies to properly evaluate data submitted for insecticide or trait registration by technology providers. While harmful effects of transgenic crops on non-target arthropods are probably relatively small compared with insecticides [11], even minor issues with experimental design could compromise the ability to detect non-target effects [2, 12].
We conducted a 2-year field experiment to estimate the optimal plot size for non-target arthropod studies in the Arizona cotton system. Our IPM program is heavily dependent on selective insecticides and the conserved biological control provided by key predators of our two key pests, Lygus hesperus Knight and whiteflies, Bemisia argentifolii Bellows and Perring (= B. tabaci MEAM1) [13–17]. We investigated how plot size impacted our ability to measure treatment differences using multiple metrics including individual predator species abundance, arthropod community structure and diversity, and biological control function.
Materials and methods
Experimental design
Experiments were conducted at the University of Arizona’s Maricopa Agricultural Center, Maricopa, AZ, United States, in 2017 and 2018. Cotton, Gossypium hirsutum L., was planted on 1 June 2017 and 4 May 2018, and grown in accordance with agronomic practices for the area. The variety planted in both years, DP1549B2XF, was a Bollgard II® XtendFlex® variety (Bayer Company, MO, USA), which provides resistance to lepidopteran insects and tolerance to dicamba, glyphosate and glufosinate herbicides.
A randomized complete block design was used in both years. Plots were established in a single field site about 3 ha in size, subdivided into four blocks. Within blocks, treatments were randomly assigned to plots.
Experimental treatments consisted of plot sizes and insecticides in a full factorial. Three square plot sizes were established: a “small” plot: 12 m long by 12 m (12 rows; 144 m2); “medium” plot: 18 m long by 18 m (18 rows; 324 m2); “large” plot: 24 m long by 24 m (24 rows; 576 m2) with ca. 1 m row spacing, and 3 m unplanted alleys. Two controls were established, a negative control, an untreated check (water only, UTC), and a positive control (a broad-spectrum insecticide) with known negative effects on the arthropod community [17] This positive control enabled us to assess the ability of our experimental design to detect an expected effect on the arthropod community, and was implemented as acephate (Orthene® 97P, Amvac Chemical Corporation, California, USA) at 1120 g a.i./ha. Acephate has commercial activity against L. hesperus, but essentially no effect on B. argentifolii. A third treatment consisted of a selective insecticide that targets B. argentifolii and with proven non-effects on the arthropod community [17]. The selective insecticide was flupyradifurone (Sivanto™ 200SL, Bayer Crop Science, North Carolina, USA) applied at 202 g a.i./ha.
The trial was sprayed with a six row (6 m) tractor-mounted boom sprayer (TJ69-8003VS TeeJet® spray tips, two over the top nozzles per row) at a volume of 112.5 l/ha. To avoid drift, insecticides were sprayed directly to plots during calm weather conditions, using low spray boom heights and reduced sprayer ground speed.
Calendar sprays with acephate and flupyradifurone treatments were made every 14 days for a total of 3 sprays at their highest labeled rates during the flowering period. These were scheduled sprays for the purposes of non-target evaluations, and therefore they were not based on needs dictated by action thresholds for pest control. Spray dates were 1 August, 15 August and 29 August in 2017, and 2 August, 16 August and 30 August in 2018. A plant growth regulator (mepiquat pentaborate, Pentia™ 99 g ai / l, BASF, Texas, USA) was sprayed in 2017 according to cotton commercial guidelines to manage the balance between vegetative and reproductive growth for cotton production. These plant growth regulator sprays were not necessary in 2018.
“Maintenance sprays” for prey uniformity
Ideally, pest levels should be comparable among insecticidal treatments as pests can change plot conditions (i.e., flower loss due to Lygus damage or excess honeydew due to whiteflies) that might affect arthropod dynamics, thereby causing gross changes across treatments that could potentially mask the effects of the intended treatments on arthropods [17]. Bordini et al. [17] deployed “maintenance sprays” when testing the selectivity of insecticides that controlled Lygus or whiteflies. The objective of these “maintenance sprays” was to preserve prey parity (L. hesperus and B. argentifolii) as much as possible among insecticidal treatments by spraying other insecticides that selectively targeted either whiteflies or Lygus [17]. Since insecticidal treatments provide control of different pests and one treatment is broad-spectrum (acephate controls L. hesperus) and the other one is selective (flupyradifurone controls B. argentifolii), we anticipated that there would be disparities among major arthropod prey. Thus, we tried to maintain comparable levels of B. argentifolii and L. hesperus in all plots as much as possible with these “maintenance sprays”. The most convenient metric for those levels is the action threshold for each pest. Thus, these sprays targeted L. hesperus and B. argentifolii, and were deployed at economic threshold levels based on standard sampling methods for these pests [18, 19].
Maintenance sprays for L. hesperus control were deployed in the untreated check and flupyradifurone treatments, but not in the positive control as acephate has commercial activity against Lygus (S1 Table in S1 File). The maintenance sprays for L. hesperus control were a selective insecticide (flonicamid, Carbine® 50WG, 98 g ai/ha, FMC Corporation, Pennsylvania, USA) applied twice in 2017 (10 and 24 August) and three times (2, 16, 30 August) in 2018. Maintenance sprays for B. argentifolii control were deployed once in the positive control and once in the untreated check in both years. We sprayed the selective insecticides, pyriproxyfen (Knack® 0.86EC, 75 g ai/ha, Valent, California, USA) on 24 August 2017, and buprofezin (Courier® 3.6SC, 390 g ai/ha, Nichino America, Delaware, USA) on 30 August 2018. As flupyradifurone has whitefly activity, no maintenance sprays were required against whiteflies; density levels there never exceeded the threshold.
Arthropod sampling
Lygus (nymphs and adults) and other arthropods were sampled concurrently with a standard 38 cm diameter sweep net. Sampling was done at three, seven and 13 days after each spray (a total of 9 weekly dates over the season each year). Twenty-five sweeps per plot were used in the small plot, 2 sets of 25 sweeps (50 sweeps) in medium plots, and 4 sets of 25 sweeps (100 sweeps) in large plots. This intensity of sampling was similar across different plot sizes to ensure similar removal of arthropods. All data were standardized to 100 sweeps, because this is the unit of measurement used for L. hesperus and predator sampling in our system [16, 18].
Densities of B. argentifolii were sampled at three and seven days after each spray, for a total of six samples each year. Ten leaves from the fifth mainstem node below the terminal per plot were randomly selected to estimate adult density and collected to estimate egg and nymph density in the laboratory. Adult density was estimated by counting individuals on the underside of leaves in situ [19]. Nymph and egg densities were estimated by counting individuals in the laboratory under magnification on a 3.88 cm2 disk taken from these leaves [20, 21].
Densities of 27 additional arthropod taxa were estimated, including key arthropod predators in our system (Vandervoet et al., 2018); Collops quadrimaculatus (Fabricius), C. vittatus (Say), Orius tristicolor (White), Geocoris punctipes (Say), G. pallens Stål, Misumenops celer (Hentz), Drapetis nr. divergens Loew and Chrysoperla carnea s.l. (Stephens). Samples were frozen and later counted in the laboratory using a dissecting microscope. These are all generalist predators. Collops spp. are soft-winged flower beetles from the family Melyridae. The hemipterans Geocoris spp. (family: Geocoridae) and Orius spp. (family: Anthocoridae) are piercing-sucking predators. M. celer are commonly known as “crab spiders” and belong to the family Thomisidae. C. carnea is a species of green lacewing from the Chrysopidae family.
Biological control function
Predator to prey ratios.
The key predators mentioned above were used to calculate predator to prey ratios as the quotient of each species per 100 sweeps to the number of B. argentifolii adults per leaf or large nymphs per leaf disc. We estimated eight predator to prey ratios as follow: M. celer/B. argentifolii adults, M. celer/B. argentifolii large nymphs, D. nr divergens/B. argentifolii adults, D. nr divergens/B. argentifolii large nymphs, O. tristicolor/B. argentifolii adults, C. carnea larvae/B. argentifolii adults, Collops spp./B. argentifolii large nymphs and G. punctipes/B. argentifolii large nymphs. For Geocoris spp., evaluations were done only with G. punctipes because G. pallens Stål densities were very low throughout the years of these trials. We chose these ratios because at certain levels they independently indicate functioning whitefly biological control in our system [16, 17, 22, 23]. We also calculated the proportion of time that these ratios were at or above functioning biological control over the season [17].
Sentinel prey.
A novel in situ sentinel prey method was developed to measure biological control function along with other sources of in-field mortality for B. argentifolii eggs and fourth instar nymphs based on procedures developed for life tables of sessile insects [24, 25]. Whiteflies are convenient and realistic as sentinel prey, because nymphs and eggs are immobile on leaves, abundant in the field, and natural enemies readily feed on them. They also are the most important pest subjected to biological control in our system. We used fourth instar nymphs because they 1) are easily seen in the field, 2) are the last instars prior to adult emergence and thus it is possible to distinguish successful emergence of adults from marks of mortality, 3) mortality is greatest during the fourth stadium, followed by mortality during the egg stage [24], 4) insecticide thresholds in our system are based on large nymph numbers in addition to adults, and 5) the majority of our primary natural enemies feed on nymphs. We used eggs because this stage is subject to the second highest mortality level, and some of our key predators (O. tristicolor; Geocoris spp. and C. carnea larvae) feed on whitefly eggs [24].
Three cohorts of at least 25 newly-laid live eggs (< 1 day old) per plot were established on 1, 16 and 29 August in 2018. The first two cohorts were established the same day of the sprays, and the third cohort was established the day before the last spray. Two cohorts of at least 25 newly molted fourth instar live nymphs, appearing flat and translucent, per plot were established on 2 August in three replicate plots, and 29 August in all plots in 2018. The first cohort was established the same day of the first spray, and the second cohort was established the day before the last spray.
Leaves with nymphs were collected 4–5 days and leaves with eggs at 7 days after establishment and taken to the laboratory for inspection. These intervals correspond with developmental times for these stages in a typical Arizona summer. Eggs and nymphs were examined to determine in-field mortality sources using a dissecting scope in the laboratory. A single observer made all determinations within a single block. Mortality was recorded as due to predation, parasitism, inviability or eggs that failed to hatch, dislodgement and unknown. Dislodgement can be due to weather like rain and/or wind or to chewing predation by beetles mainly from Collops spp. or Coccinelidae. This type of predation usually removes all trace of the individual. However, in rare cases we could observe partial nymphal cadavers or egg pedicels still anchored on leaves [24]. Predation was mainly due to the sucking predators Geocoris spp., Orius spp. and C. carnea larvae as shown in previous studies [24]. These predators evacuate the content of nymphs and eggs, leaving a transparent nymph cuticle or egg chorion on the leaf.
Different factors that affect B. argentifolii mortality occur simultaneously and so there is no obvious temporal sequence of mortality. Thus, mortality from one factor can conceal the action of a previous factor [24]. To account for this and estimate mortality accurately, marginal mortality rates were calculated for each factor ([25–28]; see [24] for formulas).
Diversity indices
Arthropods (43 taxa) collected from sweep samples were used to calculate Species Richness (S), the Shannon-Wiener Diversity Index (H), the Effective Number of Species associated with the Shannon Diversity Index (ENS) and the Shannon Evenness Index (J). These metrics were calculated for every treatment plot.
Statistical analyses
We used a mixed-model, repeated measures analysis of variance (JMP® Pro 14.2, SAS Institute Inc., Cary, NC) to test for treatment differences (insecticide and plot size) affecting the abundance of the six key predators and key pests over the season in both years. We also used this model to test for treatment differences in diversity metrics. The model included fixed effects of plot size, insecticide, year and sampling date (repeated measure) and their interactions. Block and associated interaction terms were considered random effects. The covariance structure used was AR(1). We used the proportion of maximum scaling method (POMS), which transforms each scale (predators’ scale in each year) to a common metric running from 0 (= minimum possible) to 1 (= maximum possible) [29], to minimize expected year effects in abundance of the predatory arthropods and to enable more consistent comparisons between years. We compared the mean weekly abundance of the six key predators, pests, diversity metrics using Dunnett’s test within each year. The untreated check for large plot size was the standard.
Analyses were done for all sample dates after the first application of insecticides. Before insecticidal application, pre-counts of arthropod densities were not statistically different. To facilitate the visualization of patterns, we graphically represented pest and predator abundance using cumulative arthropod-days over the season using the trapezoidal rule [30].
We calculated the proportion of dates that each key predator to prey ratio was above functioning biological control levels in our system [17]. These data were analyzed using a mixed-model that included fixed effects of insecticide treatment and year; the block variable and associated interaction terms were entered as random effects (JMP® Pro 14.2, SAS Institute Inc., Cary, NC). The same model was used to test for treatment differences (insecticide and plot size) in marginal mortality for nymphs and eggs. In all cases, Dunnett’s test was used to compare results from treatments with the large plot untreated check.
We examined the main effects of plot size on the entire arthropod predator community through Principal Response Curves (PRC), a time-dependent, multivariate analysis that depicts arthropod community trends over time for each treatment relative to a control [31–34]. We examined the main plot size effects using the large plot size as the standard. PRCs use a distribution-free F type test based on sample permutation to test for statistical significance in patterns. Principal response analyses were done in CANOCO v5 (Microcomputer Power, Ithaca, NY, USA).
Finally, we estimated Hedge’s d effect sizes between means from arthropod abundance (key predators, Lygus spp. and B. argentifolii) in each insecticidal treatment (positive control and flupyradifurone) relative to means in the untreated check for each plot size.
Results
Individual predator abundance
Plot size and its interactions were not significant for key predator abundance (P > 0.05; Table 1; Fig 1). Predator abundance in the flupyradifurone treatment was either statistically greater or not different from the UTC most of the time, being lower than the UTC in only two instances (once for O. tristicolor after the first spray in 2017, and once for M. celer after the third spray in 2018). The negative control, acephate, supported much lower predator densities as expected. The results from the negative control provide strong evidence that the experimental design was robust enough to measure the known destructive effect of this insecticide.
Asterisks correspond to treatment means for the main effect of insecticides that were significantly different from the untreated check by Dunnett’s, P < 0.05, by date and year; plot size and its interactions were not significant.
See Fig 1 for plots of density over the season.
Non-target arthropod community dynamics
The Principal Response Curve (PRC) depicts the effect of small and medium plot sizes relative to the large plot size for the untreated check, positive control and flupyradifurone separately (Fig 2). PRCs based on the first axis of redundancy analyses were not significant (P > 0.05) in any comparison (Fig 2). We also examined each insecticide compared to the untreated check for each plot size to further understand community effects (S1-S4 Figs in S1 File). As expected, the arthropod community was significantly reduced in the negative control relative to the UTC in both years and in all plot sizes (P < 0.05) (S1-S3 Figs in S1 File). Again, this validates our experimental design because we were able to clearly measure the known destructive impacts of the negative control, acephate, on the arthropod community in all plot sizes.
Principal response curves (PRC) showing plot size effects on the arthropod community for the untreated check (A), the positive control acephate (B) and flupyradifurone (C) relative to the large plot size (y = 0 line) during two growing seasons in Maricopa, AZ. The P-values denotes the significance of the PRC analysis in comparison with the large plot size over all dates based on an F-type permutation test. The product of the species weight and the canonical coefficient for a given plot size and time estimates the natural log change in density of that species relative to the large plot size. The greater the species weight the more the response for that species resembles the PRC. Negative weights indicate an opposite pattern, and weights between −0.5 and 0.5 indicate a weak response or a response unrelated to the PRC.
Whitefly and Lygus target pest abundance
Plot size and its interactions were not significant for B. argentifolii adults and large nymphs (P > 0.05) (S2 Table in S1 File). The main effect of plot size was not significant for B. argentifolii small nymphs and eggs. However, the interaction plot size*year*date was significant for small nymphs, and the same interaction along with insecticide*plot size*date were significant for eggs (P < 0.05). These are not ecologically relevant interactions as temporal effects were expected due to yearly and weekly changes in abundance, and the second interaction was due to only one contrast over twelve dates (S5 Fig in S1 File).
Plot size and its interactions were not significant for L. hesperus nymph abundance (P > 0.05) (S3 Table in S1 File). The main effect of plot size was significant for Lygus adult abundance only (P < 0.05), where adult abundance was significantly lower in the small plot compared with the large plot (P < 0.05); however, the maximum differences in abundance between plot sizes were 2–5 adults per 100 sweeps. Adults ranged from about 4 to 72 adults per 100 sweeps over the course of 18 sampling dates in this two-year study. Given this and the fact that we were able to detect insecticide effects on Lygus in small plot sizes (insecticide*plot size was not significant), the statistically significant difference identified between adult abundance in large compared to small plots is not likely to be ecologically important.
The abundance of B. argentifolii was significantly higher in the positive control than the UTC for several instances in both years across all life stages (Fig 3; S6 Fig in S1 File). The abundance of B. argentifolii was significantly lower in the flupyradifurone treatment than the UTC in as many as four instances in 2018 and three instances in 2017 (Fig 3; S6 Fig in S1 File).
Cumulative mean insect-days (error bars = S.E.) for B. argentifolii, expressed as number of adults per leaf (A), large (B), small (C) nymphs and (D) eggs per 3.88 cm2 leaf disc. Asterisks correspond to treatment means for the main effect of insecticides that were significantly different from the untreated check by Dunnett’s, P < 0.05, by date and year.
Population densities of L. hesperus were significantly lower in the positive control compared with the UTC on several dates in 2017 (mainly nymphs) and only in one dates in 2018 (Fig 4). Population densities of this pest was significantly higher in the flupyradifurone than the UTC in only one date in both years (Fig 4).
Cumulative mean insect-days for Lygus hesperus adults (A) and nymphs (B) per 100 sweeps. Asterisks correspond to treatment means for the main effect of insecticides that were significantly different from the untreated check by Dunnett’s, P < 0.05, by date and year.
Biological control function
Predator to prey ratios.
We estimated the proportion of sample dates in which key predator to whitefly ratios were at or above levels indicative of functioning biological control in our system. Plot size and its interactions were not significant (P > 0.05; Table 2). The positive control performed as expected. The proportion of time at or above functioning biological control levels of multiple ratios were significantly lower in the positive control compared to the UTC (Fig 5). This metric in the flupyradifurone treatment was significantly higher or not different from the UTC, with the exception of the ratio C. carnea larvae to B. argentifolii adult, in which the proportion of times was significantly lower than the UTC (Fig 5).
For each insecticide main effect, these proportions were compared with the UTC by Dunnett’s (*, P < 0.05) each year; plot size and its interactions were not significant.
Sentinel prey.
Neither plot size nor any of its interactions were significant for any mortality factor in the sentinel prey study with whitefly eggs or nymphs (P > 0.05; Tables 3 & 4). Total mortality, marginal predation and parasitism (nymphs only) of nymphs and eggs were significantly lower in the positive control than the UTC as expected (Figs 6 & 7).
B. argentifolii are convenient and realistic as sentinel prey, because eggs are immobile on leaves, abundant in the field, and natural enemies readily feed on them. Eggs of B. argentifolii are one of the most important life stages subjected to biological control in our system. Asterisks denote main effects for insecticides that were significantly different from the untreated check by Dunnett’s, P < 0.05; plot size and its interactions were not significant.
B. argentifolii are convenient and realistic as sentinel prey, because nymphs are immobile on leaves, abundant in the field, and natural enemies readily feed on them. Nymphs of B. argentifolii are one of the most important life stages subjected to biological control in our system. Asterisks correspond to treatment means for acephate that were significantly different from the untreated check by Dunnett’s, P < 0.05; plot size and its interactions were not significant. Marginal predation in the acephate treatment was significantly different from the untreated check only in the first nymph cohort by Dunnett’s, P < 0.05.
Diversity indices.
None of the diversity indices were significantly affected by plot size or any of its interactions (P > 0.05; S4 Table in S1 File). There also were no strong patterns related to plot size for any of the indices (Fig 8; S5 Table in S1 File).
Neither plot size nor any of its interactions were significant (P > 0.05).
Effect size.
Effect size of predator abundance between the positive control acephate or flupyradifurone and the untreated check for each plot size ranged from small (0.2), medium (0.5) to large (0.8) effects sizes using Cohen’s scale [35]. However, there were no patterns related with plot size (Table 5). We also estimated effect size for pest abundance, and there were again no patterns related with plot size (S6 Table in S1 File).
Only density is presented for the UTC.
Discussion
We investigated how plot size impacts the estimation of treatment effects relative to density, diversity and biological control function for arthropod taxa with a wide range of mobility (Arachnids, Coleoptera, Hemiptera, Diptera, etc.). We found no effect of plot size and concluded that “small” square plots (144 m2) are sufficiently large to measure insecticidal effects on non-target arthropods in Arizona cotton. We were able to clearly measure the known destructive effects of the positive control (acephate), and detected effects of the selective insecticide flupyradifurone in small plots, including improved predation and arthropod abundance. Small plots supported arthropods for all parameters compared to medium and large plots based on several metrics: 1) individual species abundance and effect sizes of key predators and pests; 2) community structure (PRCs) and diversity (indices); and 3) biological control function (mortality of sentinel prey) and success (predator to prey ratios).
The use of maintenance sprays in this study were necessary because of the potential differences in prey density due to the three treatment sprays and the different spectrum of control of the selective maintenance insecticide and the positive control [17]. The selective insecticide flupyradifurone has excellent whitefly control, and the positive control acephate is effective against Lygus and destructive to whitefly predators, releasing this pest from biological control. As a result, imbalances in prey level and in plot characteristics (i.e., boll and flower loss due to Lygus and excess honeydew due to whiteflies) could occur and might impact arthropod distribution differently in each insecticidal treatment. However, despite our best efforts to achieve prey parity with maintenance sprays for whiteflies and Lygus, there were still significant differences in prey levels in comparison with the negative control in a few instances. Even with variable abundance of prey among insecticidal treatments that had the potential to impact arthropod distribution in a few cases, results from all tested parameters were consistent across plot sizes, suggesting that our findings are robust to uncontrollable ecological variables.
Previous plot size research done in Texas cotton found that “small” plots were sufficiently large for non-target arthropod studies. Harding et al. [36] studied the effect of plot size on pests and predatory arthropods in cotton plots ranging from ca. 200 to 400 m2 treated with aerial sprays, and found that 200 m2 plots supported similar abundance of prey and predators compared to larger plots. Their 200 m2 plot was only three rows larger than our smallest plot size. However, Harding et al. [36] only sprayed the trial with broad-spectrum insecticides, sampled with a D-vac machine, analyzed arthropods all together, did not report the use of “maintenance sprays” to attempt prey parity among treatments. Further, L. hesperus, a significant pest in the United States, was not the object of their study at that time. Besides these differences, our study was more robust by also measuring additional important ecological variables such as key predator to whitefly ratios indicating functional biological control in our system along with direct measurements of predation (sentinel prey), and effects on community structure and diversity that included 43 taxa.
Plot size in non-target arthropod studies varies considerably [3]. For example, studies involving non-target species in arable crops used plots ranging from 11,800 to 16,900 m2 in corn [37], and some recent studies used plots of 12 m2 in soybeans [38], 21 m2 in quinoa [39] and 170 and 500 m2 in cotton [40, 41]. In meta-analyses to examine the non-target impacts of transgenic Bt crops, Wolfenbarger et al. [42] found no pattern in effect sizes between Bt and non-Bt crops over plot sizes ranging from 20–175,000 m2. This finding was recently supported by a comprehensive meta-analysis of Bt maize [43]. In contrast, studies conducted in cereals using a variety of plot sizes suggested that small plot trials may underestimate potential harmful effects of insecticides on non-target arthropods [4–6].
Our small plot here was the smallest practical plot that we could use without compromising sample unit size and over-sampling the fauna in our system. In our small plots, almost the entire length of two middle rows is needed to complete 25 sweeps and still avoid sampling the plot edges. It also was important to have a significant number of interior rows designated for sampling so that one could use alternate rows over time and minimize plant damage caused by sweeps. The minimal plot size is also a factor of the sampling method. Thus, systems that use other sampling methods (i.e., beat sheets) could potentially be subjected to smaller plots whereas other sampling methods might dictate the need for a larger plot. Plot geometry also needs to be considered. Our plots were square, however; long rectangular plots of the same area might not perform well because the distance between field edges of the longest side of rectangular plots would be reduced, which could influence arthropod movement and behavior differently. Distance between plots, plot isolation by alternate crops or bare ground and surrounding vegetation can impact arthropod movement as well [2]. We did not explicitly examine the effects of interplot space, but our results from the small plot size suggest that interplot movement across 3 meters of bare ground at high air temperatures in the summer was inconsequential because we were able to detect treatment differences in these small plots. Ground dwelling arthropods might behave differently, but they were not directly the subject of this investigation.
At the other end of the spectrum, there were practical constraints to testing even larger plot sizes, including significant increases in costs of land rental, irrigation costs, labor, etc. Previous research done in larger plots in Arizona cotton found no effect of plot size in a long-term study testing the effects of transgenic Bt cotton on non-target arthropod abundance between plots ranging from 1200 m2 to 20,000 m2 [11]. Larger plots are normally more heterogenous and might require greater sampling effort or more replicates to detect changes in population density [11]. Another factor that needs to be considered is the potential loss of homogeneity of blocks as larger plots are used. Blocking field experiments control for nuisance variations like irrigation, soil texture, adjacent fields; however, at some point, inter-plot differences risk being larger than block differences or blocks cannot be efficiently established.
This study investigated the effect of plot size on several taxa with a wide range of mobility in Arizona cotton. We demonstrated that “small” square plots (144 m2) are sufficiently large for non-target arthropod studies in Arizona cotton. Small plots allowed us to detect treatment differences, and supported similar individual predator abundance, arthropod community structure and diversity, and biological control function and success compared with larger plot sizes. Therefore, our results are applicable to cotton systems with related pests, predators or other fauna with similar dispersal characteristics. These findings can be applied to target and non-target organism studies in cotton, including studies involving transgenic cotton with insecticidal properties. Moreover, these results also might be generalizable to other crop systems with similar fauna.
This new information should be helpful to growers, researchers, technology providers and regulatory agencies in measuring impacts of various insect control technologies on non-target arthropods in cotton. Furthermore, they point to a scale of testing that should be considered when developing any IPM guidelines that are provided to farmers for use under commercial conditions. These results will guide Arizona’s field evaluation of current and future technologies with goals of providing reliable information about risk to non-target arthropods to our growers [44].
Supporting information
S1 File. Supplemental information contains all the supporting tables and figures.
https://doi.org/10.1371/journal.pone.0272831.s001
(PDF)
Acknowledgments
We thank F. Bojorquez, G. Lizarraga, J. Partida, A. Brown, E. Thacker, T. Thacker, N. Pier, P. Merten, M. Cruz, F. Pat, J. Fan and J. Trejo for their lab technical assistance. Thanks to P. Asiimwe for reviewing an earlier draft of this manuscript and providing helpful comments. We thank the pest control advisers Tom Montoya, Nathan Kempton, Ryan Tregaskes for supporting our proposal for this project. We thank the Entomology and Insect Science Graduate Interdisciplinary Program at University of Arizona. Mention of trade names or commercial products in this article is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture. USDA is an equal opportunity provider and employer.
References
- 1. Jepson P, Thacker J. Analysis of the spatial component of pesticide side-effects on non-target invertebrate populations and its relevance to hazard analysis. Functional Ecology. 1990:349–55.
- 2. Prasifka JR, Hellmich RL, Dively G, Lewis LC. Assessing the effects of pest management on nontarget arthropods: the influence of plot size and isolation. Environmental Entomology. 2005;34(5):1181–92.
- 3. Macfadyen S, Banks JE, Stark JD, Davies AP. Using semifield studies to examine the effects of pesticides on mobile terrestrial invertebrates. Annual Review of Entomology. 2014;59:383–404. pmid:24160417
- 4. Pullen A, Jepson P, Sotherton N. Terrestrial non-target invertebrates and the autumn application of synthetic pyrethroids: experimental methodology and the trade-off between replication and plot size. Archives of Environmental Contamination and Toxicology. 1992;23:246–58.
- 5. Duffield S, Aebischer N. The effect of spatial scale of treatment with dimethoate on invertebrate population recovery in winter wheat. Journal of Applied Ecology. 1994:263–81.
- 6. Kennedy P, Conrad K, Perry J, Powell D, Aegerter J, Todd A, et al. Comparison of two field-scale approaches for the study of effects of insecticides on polyphagous predators in cereals. Applied Soil Ecology. 2001;17(3):253–66.
- 7. Alix A, Bakker F, Barrett K, Brühl C, Coulson M, Hoy S, et al., editors. ESCORT 3: Linking non-target arthropod testing and risk assessment with protection goals. Proceedings of the European Standard Characteristics Of non-target arthropod Regulatory Testing workshop ESCORT; 2012.
- 8. Furlong MJ, Zalucki MP. Exploiting predators for pest management: the need for sound ecological assessment. Entomologia Experimentalis et Applicata. 2010;135(3):225–36.
- 9. Zalucki MP, Furlong MJ, Schellhorn NA, Macfadyen S, Davies AP. Assessing the impact of natural enemies in agroecosystems: toward" real" IPM or in quest of the Holy Grail? Insect Science. 2015;22(1):1–5. pmid:25205315
- 10. Macfadyen S, Davies AP, Zalucki MP. Assessing the impact of arthropod natural enemies on crop pests at the field scale. Insect Science. 2015;22(1):20–34. pmid:25219624
- 11. Naranjo SE. Long-term assessment of the effects of transgenic Bt cotton on the abundance of nontarget arthropod natural enemies. Environmental entomology. 2005;34(5):1193–210.
- 12. Wold SJ, Burkness EC, Hutchison W, Venette RC. In-field monitoring of beneficial insect populations in transgenic corn expressing a Bacillus thuringiensis toxin. Journal of Entomological Science. 2001;36(2):177–87.
- 13. Ellsworth PC, Martinez-Carrillo JL. IPM for Bemisia tabaci: a case study from North America. Crop protection. 2001;20(9):853–69.
- 14. Naranjo SE, Ellsworth PC. Fifty years of the integrated control concept: moving the model and implementation forward in Arizona. Pest Management Science: formerly Pesticide Science. 2009;65(12):1267–86.
- 15.
Ellsworth P, Fournier A, Frisvold G, Naranjo S, editors. Chronicling the socio-economic impact of integrating biological control, technology, and knowledge over 25 years of IPM in Arizona. Proceedings of the 5th International Symposium on Biological Control of Arthropods, Langkawi, Malaysia, September 11–15, 2017; 2017: CABI Wallingford UK. https://acis.cals.arizona.edu/agricultural-ipm/vegetables/vegetable-outputs/publications/publications-view/chronicling-the-socio-economic-impact-of-integrating-biological-control-technology-and-knowledge-over-25-years-of-ipm-in-arizona [accessed 4 April 2022].
- 16. Vandervoet TF, Ellsworth PC, Carrière Y, Naranjo SE. Quantifying conservation biological control for management of Bemisia tabaci (Hemiptera: Aleyrodidae) in cotton. Journal of Economic Entomology. 2018;111(3):1056–68. pmid:29546368
- 17. Bordini I, Ellsworth PC, Naranjo SE, Fournier A. Novel insecticides and generalist predators support conservation biological control in cotton. Biological Control. 2021;154:104502.
- 18.
Ellsworth PC, Barkley V. Cost-effective Lygus management in Arizona cotton. In: Silvertooth J.C. (ed.), Cotton: A College of Agriculture and Life Sciences. University of Arizona, College of Agriculture and Life Sciences Tucson, AZ. 2001; 299–307. https://repository.arizona.edu/handle/10150/211330 [accessed 4 April 2022].
- 19. Ellsworth P, Palumbo JC, Naranjo SE, Dennehy TJ, Nichols RL. Whitefly management in Arizona cotton 2006. 2006.
- 20. Naranjo S, Flint H, Henneberry T. Spatial distribution of adult Bemisia tabaci in cotton and development and validation of fixed-precision sequential sampling plans for estimating population density. Environ Entomol. 1995;24:261–70.
- 21. Naranjo SE, Flint HM. Spatial distribution of preimaginal Bemisia tabaci (Homoptera: Aleyrodidae) in cotton and development of fixed-precision sequential sampling plans. Environmental Entomology. 1994;23(2):254–66.
- 22. Ellsworth P.C., Pier N., Fournier A., Naranjo S.E. (2019a). Making use of predators in cotton. Arizona Pest Management Center field crops IPM shorts. University of Arizona, Cooperative of Extension. 2019a. https://cals.arizona.edu/crops/cotton/files/PtoPlaminate.pdf [accessed 4 April 2022].
- 23.
Ellsworth P.C., Pier N., Fournier A., Naranjo S.E., Vandervoet T. Predator “thresholds”. Arizona Pest Management Center field crops IPM Shorts. University of Arizona, Cooperative of Extension. 2019b. https://acis.cals.arizona.edu/docs/default-source/ipm-shorts/wfbit.pdf?sfvrsn=9e760083_2 [accessed 4 April 2022].
- 24. Naranjo SE, Ellsworth PC. Mortality dynamics and population regulation in Bemisia tabaci. Entomologia experimentalis et Applicata. 2005;116(2):93–108.
- 25. Naranjo SE, Ellsworth PC. Methodology for developing life tables for sessile insects in the field using the whitefly, Bemisia tabaci, in cotton as a model system. JoVE (Journal of Visualized Experiments). 2017;(129):e56150.
- 26. Royama T. Evaluation of mortality factors in insect life table analysis. Ecological Monographs. 1981;51(4):495–505.
- 27. Buonaccorsi JP, Elkinton JS. Estimation of contemporaneous mortality factors. Researches on Population Ecology. 1990;32(1):151–71.
- 28. Elkinton JS, Buonaccorsi J, Bellows T Jr, Van Driesche R. Marginal attack rate, k‐values and density dependence in the analysis of contemporaneous mortality factors. Population Ecology. 1992;34(1):29–44.
- 29. Moeller J. A word on standardization in longitudinal studies: don’t. Frontiers Media SA; 2015. p. 1389. pmid:26441764
- 30. Ruppel RF. Cumulative insect-days as an index of crop protection. Journal of economic entomology. 1983;76(2):375–7.
- 31. Van den Brink PJ, Ter Braak CJ. Multivariate analysis of stress in experimental ecosystems by principal response curves and similarity analysis. Aquatic Ecology. 1998;32:163–78.
- 32. Van den Brink PJ, Braak CJT. Principal response curves: Analysis of time‐dependent multivariate responses of biological community to stress. Environmental Toxicology and Chemistry: An International Journal. 1999;18(2):138–48.
- 33.
Ct Braak, Smilauer P. CANOCO Reference manual and user’s guide to Canoco for Windows: software for Canonical Community Ordination (version 4). Centre for Biometry. Microcomputer Power, Ithaca, NY. 1998.
- 34.
Ter Braak CJ, Šmilauer P. Canoco reference manual and user’s guide: software for ordination, version 5.0. Microcomputer Power, Ithaca, NY. 2012.
- 35.
Cohen J. Statistical power analysis for the behavioral sciences: Routledge; 2013.
- 36. Harding JA, Dupnik TD, Wolfenbarger DA, Fuchs TW. Evaluation of plot size for arthropod studies in cotton. Texas A&M tech. bull. 1975;1–10.
- 37. Candolfi M, Brown K, Grimm C, Reber B, Schmidli H. A faunistic approach to assess potential side-effects of genetically modified Bt-corn on non-target arthropods under field conditions. Biocontrol Science and Technology. 2004;14(2):129–70.
- 38. Amin MR, Oh SD, Park S-Y, Ha K, Kang S, Park J-H, et al. The effect of thioredoxin-gene-expressed transgenic soybean on associated non-target insects and arachnids. Plant Biotechnology Reports. 2022:1–12.
- 39. Cruces L, de la Peña E, De Clercq P. Field evaluation of cypermethrin, imidacloprid, teflubenzuron and emamectin benzoate against pests of quinoa (Chenopodium quinoa Willd.) and their side effects on non-target species. Plants. 2021;10(9):1788. pmid:34579320
- 40. Marques LH, Lepping M, Castro BA, Santos AC, Rossetto J, Nunes MZ, et al. Field efficacy of Bt cotton containing events DAS-21023-5× DAS-24236-5× SYN-IR102-7 against lepidopteran pests and impact on the non-target arthropod community in Brazil. Plos one. 2021;16(5):e0251134.
- 41. Kaur J, Aggarwal N, Kular J. Abundance and diversity of arthropods in transgenic Bt and non-Bt cotton fields under Indian conditions. Phytoparasitica. 2021;49:61–72.
- 42. Wolfenbarger LL, Naranjo SE, Lundgren JG, Bitzer RJ, Watrud LS. Bt crop effects on functional guilds of non-target arthropods: a meta-analysis. Plos one. 2008;3(5):e2118. pmid:18461164
- 43. Meissle M, Naranjo SE, Kohl C, Riedel J, Romeis J. Does the growing of Bt maize change abundance or ecological function of non-target animals compared to the growing of non-GM maize? A systematic review protocol. Environmental Evidence. 2014;3(1):1–13.
- 44.
Bordini I, Fournier A, Naranjo SE, Pier N, Ellsworth PC. Cotton Insecticide Use Guide: Knowing and Balancing Risks. Arizona Pest Management Center field crops IPM shorts, University of Arizona, Cooperative of Extension. 2020. https://acis.cals.arizona.edu/docs/default-source/ipm-shorts/cottoninsecticiderisk.pdf?sfvrsn=a5628a8b_0 [accessed 20 April 2022].