Reducing Drosophila suzukii Management Challenges: An Alternative to Insecticide Cover Sprays

We tested border sprays in a less preferred crop, highbush blueberry (Bellamy et al., 2013).

Experimental site and design. A ~26 ha ‘Liberty’ blueberry (Vaccinium corymbosum L.) site in Albany, Oregon, managed using conventional practices, was selected during 2012 and 2013. The site was arranged in a randomized complete block design with three replications in the harvest seasons (seven to nine weeks). Each block had a randomly assigned border spray and cover spray plot (2.8 – 5.3 ha). Blocks 1 and 2 were drip irrigated and block 3 was overhead irrigated. In 2013, weed mat and trellis were installed in blocks 1 and 2 by the grower. Eight clear-cup monitoring traps of adult D. suzukii (Lee et al., 2012) were used per plot, with four placed in the crop border (~5 m into the field on each side of a plot) and four in the crop interior (40–60 m into the field from the border trap).

Monitoring D. suzukii, non-target arthropods, and fruit knockdown. Adult and larval D. suzukii were monitored by trapping and larval extraction methods, respectively, twice per week, as previously described. Non-target arthropods were collected as previously described (yellow sticky cards, leaf collections, sweep net, and vacuum). Beat sheet sampling was not used out of concern for excessive fruit knockdown. Plants near the east and west adult D. suzukii monitoring traps, representing the crop border and interior of each plot, were selected to quantify fruit knockdown from each spray strategy. Within 24 hours prior to application (border and cover sprays), fruit that dropped naturally or from hand-harvest was cleared from the area underneath the selected plant’s canopy, and within 24 hours after spray applications, fruit knocked down by sprayers was collected, weighed (g), and recorded.

Insecticide applications. Growers or crop consultants made all pesticide decisions and applications at the blueberry sites based on biweekly reports of D. suzukii adult and larval counts in the field. Table 1 shows insecticides, rates, cultivars, and mean number of cover and border spray applications per treatment plot at each site. Mean number of sprays per treatment plot versus total sprays was displayed in Table 1 because of differences in application numbers between blocks.

Border spray applications in blueberry were made with a cannon sprayer (J-1000, Jacto Inc., Pompea, Brazil) traveling around the perimeter of the crop and spraying pesticides up to 30 m into the field. Cover spray applications were made with an over-the-row Trellis Boom (Rear’s Manufacturing Company, Eugene, Oregon) sprayer that treated two entire plant rows per pass in 2012 and a track-elevated sprayer (TR4 Tracker, GK Machine Inc., Donald, Oregon) that treated four entire plant rows per pass in 2013. All insecticide applications (Table 1) targeted D. suzukii with the exception of a neonicotinoid applications to control an aphid outbreak in 2012.

Data analysis. All statistical analyses were made in R (R Core Team, 2013) using RStudio (RStudio, 2012) with α = 0.05. Cochran-Mantel-Haenszel chi-squared test for count data was used to determine if sums of each gender were in excess in 2 × 2 tables for each block (Ramsey & Schafer, 2002). Fisher’s Exact Test for count data was used for small sample sizes (Ramsey & Schafer, 2002).

The border spray trial was a completely randomized block design in a two-factor factorial experiment. There were two levels within each treatment (border and cover sprays) and trap position (border and interior traps). Linear mixed effects model fit by restricted maximum likelihood (REML) was used to detect differences in mean adults captured across the harvest season with treatment and trap position as fixed effects and treatment plots nested within blocks as random effects (larval counts were either zero or too low to permit analysis). A drop in deviance test was performed to fit the most appropriate model (Ramsey & Schafer, 2002). The model with the lowest Akaike information criterion (AIC) and non-significant difference between other models was deemed most appropriate using ANOVA. When the full model was most appropriate, linear combinations of coefficients were used to answer treatment and trap position questions. All P-values are reported with 95% confidence intervals (CI) and Bonferroni adjusted when two comparisons were made (α = 0.025). Natural log-transformation was used when normality or equal variance assumptions were not met. Log₁₀ (x + 1)-transformation was used to transform 0 when present in the dataset. Estimates and confidence intervals were back-transformed, in the case of transformed data, and presented in tables along with the original means and standard errors.

Individual non-target arthropods collected seven days post-harvest in 2012 and 2013 were averaged across collection types (sweep, vacuum, beat sheet, yellow sticky cards, and leaf) and all traps within a treatment plot to assess the overall impact of treatments. For example, anthocorids from all collection methods and trap captures within a treatment plot were combined and expressed as mean number per plot during post-harvest. This was done to detect plot level population changes, regardless of collection method, and to boost otherwise low counts.

Only sufficiently large enough counts of non-target arthropods were included in the analysis. If the mean number of a non-target arthropod was less than 1.0 per treatment plot, the arthropod was not included in the analysis because counts were deemed too low to draw any firm conclusions. Fruit knockdown collected after insecticide applications in the border spray trials were averaged across collection periods and for each trap position (i.e., fruit knockdown per trap position within a treatment plot) and analyzed as previously described for D. suzukii and non-target arthropods.