Final report for GNE22-305
Project Information
Monitoring insect abundance is essential for decision-making and thus a fundamental component of integrated pest management (IPM); however, these decisions are rarely based on the abundance of beneficial insects. To enhance IPM programs, monitoring tools for beneficial insects, such as pollinators and natural enemies of pests, need to be developed. While it is understood that beneficials are attracted to plant volatiles like methyl salicylate (MeSA), there are several considerations needed before volatiles can be used for monitoring. For instance, the composition of the landscape in which they live, such as the amount of non-crop habitats, can affect their response to plant volatiles. However, the combined effects of landscapes and local management practices on the response of beneficial insects to plant volatiles remains unknown. This research will determine how the landscape and local management of cranberry agroecosystems affect the beneficial insect community and subsequent ecosystem services in conjunction with the common and commercialized plant volatile MeSA. To achieve this, a trapping network will be established across 50 cranberry beds in the three largest cranberry farms in New Jersey. In each bed, sticky and pan traps baited with MeSA and unbaited traps will be used to monitor natural enemy and pollinator abundance throughout the season. In addition, sentinel eggs and exclusion cages will be placed to assess predation and pollination services. Results from these studies will help develop tools for monitoring beneficial insects that can be used in pest management decisions and conservation biocontrol.
- Investigate the effects of landscape (habitat composition) and local management practices (crop variety and pesticide usage) on response of beneficial insects to plant volatiles.
- Investigate the effects of landscape and local management practices on the ecosystem services provided by beneficial insects: biological control (predation) and pollination.
The purpose of this study is to investigate the effects of landscape composition and local management practices on plant volatile recruitment of beneficial insects, which have practical implications in integrated pest management (IPM) decision-making and biocontrol. Making decisions in IPM relies heavily on monitoring to see what insect pests are present. While many monitoring approaches are directed at pests, beneficial insects such as pollinators and natural enemies of pests are rarely considered in decision-making, despite them being important. Many beneficials are known to utilize volatile organic compounds from the plants, for example predators and parasitoids, to locate their host or prey (Price et al. 1980). Since the 1980s, entomologists have been researching these volatiles and subsequently utilizing synthetic plant volatiles to attract these beneficial insects to crops (Rodriguez-Saona et al. 2011). One such plant volatile is methyl salicylate (MeSA) that is commonly released by flowers and other plant parts particularly after insect feeding damage (Vlot et al. 2009). The attraction of beneficial insects to MeSA has been studied in multiple crops, including cranberries, and is known to attract many beneficial insects, such as hoverflies, lady beetles, lacewings, among others (Rodriguez-Saona et al. 2011). Therefore, MeSA can potentially be used to monitor beneficial insects and conservation biological control in agricultural crops. In fact, a lure called PredaLure that contains MeSA is commercially available for this purpose. However, for these lures to be successful, beneficial insects need to be present in the agroecosystem, which means further studying of the agroecosystem in the context of landscape composition and local management practices is necessary.
There are many studies about how landscapes and management practices interact to affect beneficial insect abundance and diversity (Tscharntke et al. 2012). For instance, hoverflies, which are predators as larvae and pollinators as adults, are more diverse and abundant in areas with higher floral diversity, meaning that with higher floral diversity, there is likely to be more predation of pests and pollination (Gervais et al. 2018). Thus, one can theorize that agroecosystems with higher diversity will also have higher amounts of beneficial insects. Moreover, diverse agroecosystems and non-crop habitats provide refuge for beneficials that would otherwise be killed by the pesticide regimes and toxic sprays commonly applied by farmers (Pandey et al. 2022). To date, no studies have studied the correlation that landscape composition and management practices have on beneficial insect attraction to plant volatiles. Understanding the effects of landscape factors and local management practices on the response of beneficial insects to synthetic plant volatiles will help develop new and efficient monitoring techniques for them, which could play into sustainable IPM strategies like conservation and augmentation biocontrol. As such, this proposed research directly connects to the Northeast SARE Outcome Statement, of researching key issues in sustainable agriculture. Monitoring beneficial insects will allow growers to implement these biocontrols into decision-making in IPM programs, which could lead to a decreased reliance on pesticide applications and thus a more environmentally and possibly economically friendlier approach to pest management.
Citations
Gervais, A., Chagnon, M., Fournier, V. (2018) Diversity and pollen loads of flower flies (Diptera: Syrphidae) in cranberry crops. Annals of the Entomological Society of America, 111(6), pp.326-334. https://doi.org/10.1093/aesa/say027
Pandey, S., Johnson, A. C., Xie, G., Gurr, G. M. (2022). Pesticide Regime Can Negate the Positive Influence of Native Vegetation Donor Habitat on Natural Enemy Abundance in Adjacent Crop Fields. Frontiers in Ecology and Evolution, 10:815162. https://doi.org/10.3389/fevo.2022.815162
Price, P. W., Bouton, C. E., Gross, P., McPheron, B. A., Thompson, J. N., Weis, A. E (1980). Interactions Among Three Trophic Levels: Influence of Plants on Interactions Between Insect Herbivores and Natural Enemies. Annual Review of Ecology and Systematics, 11, pp.41-65. https://doi.org/10.1146/annurev.es.11.110180.000353
Rodriguez-Saona, C., Kaplan, I., Braasch, J., Chinnasamy, D., Williams, L. (2011). Field responses of predaceous arthropods to methyl salicylate: A meta-analysis and case study in cranberries. Biological Control, 59(2), pp.294-303. https://doi.org/10.1016/j.biocontrol.2011.06.017
Tscharntke, T., Tylianakis, J. M., Rand, T. A., Didham, R. K., Fahrig, L., Batary, P., Bengtsson, J., Clough, Y., Crist, T. O., Dormann, C. F., Ewers, R. M., Frund, J., Holt, R. D., Holzschuh, A., Klein, A. M., Kleijn, D., Kremen, C., Landis, D. A., Laurance, W., Lindenmayer, D., Scherber, C., Sodhi, N., Steffan-Dewenter, I., Thies, C., van der Putten, W. H., Westphal, C. (2012). Landscape moderation of biodiversity patterns and processes - eight hypotheses. Biological Reviews 85, pp.661-685. https://doi.org/10.1111/j.1469-185X.2011.00216.x
Vlot, A. C., Dempsey, D. A., Klessig, D. F. (2009). Salicylic Acid, a Multifaceted Hormone to Combat Disease. Annual Review of Phytopathology, 47, pp.177-206. https://doi.org/10.1146/annurev.phyto.050908.135202
Research
1. Study system
1.1 Sites
A field experiment was conducted in 2022 and 2023 to measure the response of natural enemies to MeSA. Fifty beds (sites) across the three largest commercial cranberry farms (Pine Island Cranberry Co., JJ White Native Jersey Fruits, Cutts Bros) in New Jersey, USA were chosen using a random number generator, making sure no two beds were direct neighbors, afterwhich every site was georeferenced using a Global Positioning System (GPS). Each bed was one of five cultivars (‘Ben Lear’, ‘Crimson Queen’, ‘Early Black’, ‘Mullica Queen’, and ‘Stevens’), and there were equal numbers of beds (10) per cultivar. These cultivars were chosen because they are the most commonly grown in New Jersey. The same beds were used for both 2022 and 2023. All pesticide (insecticides, fungicides, herbicides) spray records were collected per bed per month at the end of each growing season.
1.2 Landscape analysis
To measure landscape, we utilized the most recent land use/cover map provided by the New Jersey Department of Environmental Protection. Using ArcGIS Pro, we classified 6 land cover types: agriculture (which was solely cranberries, since not a lot else is farmed in those areas), barren land, forest, urban, water, and wetland. We then uploaded the coordinates of our sites and drew four spatial buffer zones around each coordinate, at 100 m, 250 m, 500 m, and 1500 m radii. Using the ZonalMetrics toolbox for ArcGIS Pro (Adamczyk and Tiede 2017), we calculated landscape composition and configuration metrics of land covers for each buffer zone. We analyzed the Shannon diversity of land covers per zone, the total edge length between land covers per zone, the largest patch index per zone, numbers of patches of each cover per zone, and the percent area of each cover per zone.
2. Beneficial insects
2.1 Natural enemies
In order to sample the natural enemy community, at each bed we set up a pair of yellow sticky traps (23 × 28 cm Unbaited Pherocon AM; Trécé Inc., Adair, Oklahoma) at the canopy level at least 10 m away from both the edge of the bed and 20 m from each other. Cranberries are a viny ground-covering plant that grows in a mat, so we drove 80 cm metal pipes bent 90° at the 60 cm mark into the ground to hang each trap. One of the traps was baited with a MeSA lure (5 g load/lure; 90 d lure; average release rate ∼35 mg/day over a 4 week period at 30 °C constant in the lab; PredaLure, AgBio Inc.), and the other trap was unbaited (control) with a square white paper (8 × 8 cm) with a hole punch that visually looks similar to the MeSA lure. The bait and paper were hung directly adjacent to their respective yellow sticky trap without touching it. Yellow sticky traps are thought to be the best sampling methods for natural enemy communities in cranberries, and can easily be used with semiochemicals (Rodriguez-Saona et al. 2011, Rodriguez-Saona et al. 2012, Rodriguez-Saona et al. 2020). Every trap was out for one week at a time per month between May to August, allowing a snapshot of the community from prebloom to preharvest (one pair of sticky traps per bed per month May-August for two years). This time period is also when the worst of the cranberry insect pests of New Jersey are active as well (Ben-Zvi and Rodriguez-Saona 2023). The MeSA lure was replaced once a month. The traps were taken back to the lab, and all trapped arthropods were counted and identified using a stereomicroscope (Nikon SMZ-U, Tokyo, Japan). The arthropods were mostly identified to family. Some arthropods were not identified to family and were put into broader groupings, notably midges and parasitoid wasps. These arthropods were characterized as natural enemies and herbivores/ non natural enemies. Counts of some very small and numerous arthropods, like thrips and whiteflies, were estimated by averaging their counts from 4 squares across the 63 squares of the sticky trap grid.
2.2 Pollinators
To sample the pollinators, we set up white pan traps (Disposable 12 oz. White Plastic Soup Bowls, Amazon) with 150 mL water and 2 mL clear scentless soap (PROVON, Office Supplies). Because cranberries are viny and grow in a mat, balancing a full pan trap is rather difficult. Therefore, we drove a 60 cm long 1.2 cm diameter metal pipe into the ground, and glued a PVC pipe connector to the bottom of each pan trap. The pan traps with the PVC connector fit snugly on the pipe and were stable, and each trap was placed at canopy level. Each site had two pan traps 20 m apart. A white MeSA-baited sachet (5 g load/lure; 90 d lure; average release rate ∼35 mg/day over a 4 week period at 30 °C constant in the lab; PredaLure, AgBio Inc.) hung above one pan trap per site from a 80 cm long metal pipe, meanwhile the other (control) trap in the site had a white square paper that visually looked like the bait. Previous studies show that Traps were set for 48 hours at a time twice during cranberry bloom (once in June and once in July) in 2022 and 2023. Afterwards, all trapped arthropods were placed in vials with 70% ethanol and counted and identified at a later date. Every arthropod was at minimum identified to family, with several genera identified – particularly within bees and syrphids. These arthropods were characterized as potential pollinators, non-pollinating herbivores, and non-pollinating predators/parasitoids. For the sake of this study, only the potential pollinators were analyzed.
3. Ecosystem services
3.1 Predation
After the sticky traps were taken down, the MeSA lure was left in place and we set up sentinel Spodoptera frugiperda Smith (Lepidoptera: Noctuidae) egg masses in the same place. The egg masses were bought frozen from Frontier Agricultural Sciences (Delaware, USA). Spodoptera frugiperda is not a cranberry pest, however the predation of their dead egg masses can represent generalist predation ecosystem services. The egg masses arrived on a large wax paper sheet, and we cut out individual masses consisting of around 100 eggs using scalpels and scissors, making sure there was only one layer of eggs. We glued one egg mass per square green paper (8 × 8 cm) that was stapled to wooden stakes (25 cm tall, 2.5 cm wide). We set up three egg mass stakes in a circle around the MeSA bait and three around the paper bait at a 50 cm radius. The egg masses deployed for 48 hours at a time, once a month between May to August, and never overlapped with the sticky traps nor pan (three pairs of sentinel egg masses per bed month May-August for two years). We labeled the egg masses, took pictures immediately before and after the 48 hour time period in the field, and counted the eggs in all the pictures. Afterwards, we calculated predation rates by dividing the total eggs before by total eggs after.
3.2 Pollination
A field experiment investigating the effects of MeSA on pollination services was conducted in 25 of the 50 sites chosen for the pollinator study above, in 2022 and 2023. The 25 sites were randomly chosen from the 50 with a random number generator, making sure there were still an equal amount (5) of sites per cultivar.
To look at pollination we set up two cages (30 cm x 30 cm x 30 cm metal grid wrapped in a thin white fabric at the base and open at the top) in each of the chosen sites 20 m apart. These cages were open to allow pollinators to enter. One of the cages was baited with the MeSA sachet, and the other had the paper square. The cages were set up before bloom (beginning of June) and closed at the beginning of fruit set (end of July) to prevent potential herbivores. Cranberries are a pollinator dependent crop, thus yield was used as a proxy for pollination. All the formed fruits were collected from each cage at the end of the season (September) and were counted, then 10 berries per sample were measured and weighed. This experiment was repeated for 2022 and 2023 in the same location.
4. Statistical analysis
4.1 Natural enemies
4.1.1 Natural enemies with methyl salicylate
The response variables in the community ecology study were the organism count. Natural enemies and herbivores were mostly identified to family. We also further calculated total abundances and Shannon diversity of natural enemies at each sire, as well as that of herbivores at each site.
For the plant level, we began with Hellinger transforming the natural enemy count data using “decostand” function, testing for dispersion homogeneity using the “betadisper” and “permutest” functions, then running a Permutational Multiple Analysis of Variance (PERMANOVA) to look at the effects of MeSA on the natural enemy community with 999 permutations on a Bray-Curtis distances using “adonis2” function in RStudio (ver. 4.2; R Core Team 2024) all with the “vegan” package (Oksanen et al. 2025). After that, we looked at one natural enemy taxa at a time and ran generalized linear mixed models (GLMMs) on the MeSA lure with random factors of bed nested in farm and month nested in year, using the “glmer” function in the “lme4” package in RStudio (Bates et al. 2015). When looking at one natural enemy family at a time we did not look at Hellinger transformed data, rather raw data. The distributions used for the GLMMs were either Poisson, Negative Binomial or Poisson-Lognormal, and that was determined with the “overdisp_fun” function from the “PsychHelperFunctions” package (Huff and Papenmeier 2022). We also then looked at the effect of MeSA with the same random factors on the total abundance and Shannon diversity of natural enemies using a Poisson-Lognormal distributed GLMM and a Gamma log link distributed GLMM respectively. Afterwards, we repeated this same exact process except with the herbivore count data.
4.1.2 Natural enemies and local management
Our local management variables consisted of the number of insecticides sprayed per bed, the number of fungicides sprayed per bed, the number of herbicides sprayed per bed, and the cultivar per bed. Our response variables were still the natural enemy counts and random factors were still bed nested in farm and month nested in year, except now we were looking at each local management factor, and the interactions of cultivar and each pesticide amount individually as fixed factors. We similarly ran GLMMs that either had Poisson, Negative Binomial, or Poisson-Lognormal distributions for each natural enemy taxon, and we now chose best fit models by eliminating insignificant fixed variables and interactions until we achieved the lowest Akaike’s Information Criterion (AIC) value. The AIC value does not inherently mean anything, rather it is used in model comparison and selection (Symonds et al. 2011). Other than just knowing how management affects the natural enemies, we also wanted to know how MeSA is involved. Therefore, we ran another round of GLMMs for every natural enemy taxon with the interaction of the MeSA lure and each of the local management variables as explanatory factors and the same random factors as before. Again we removed insignificant interactions and variables as per the AIC value.
4.1.3 Landscape level natural enemies
Lastly for the landscape level, the explanatory variables were the all landscape factors per buffer zone (Shannon diversity of land covers, the total edge length between land covers, the largest patch index, numbers of patches of each land cover, and the percent area of each land cover), year, and the random site nested in farm. We ran a partial redundancy analysis (RDA) with the “rda” function in the “vegan” package (Oksanen et al. 2025), which is similar to a principal component analysis, but it considers a constrained environment, i.e the landscape variables. It was a partial RDA because we only wanted to account for the landscape while controlling the year and month, so we added those as covariates. For this analysis we looked at each buffer zone separately. Before running the partial RDA, we standardized the numeric explanatory variables to have a mean of 0 and standard deviation of 1, as well as the community response with the Hellinger transformation, both by utilizing the “decostand” function. We ran several rounds of RDA and performed forward selection on variables using the “ordiR2step” function from the “vegan” package (Oksanen et al. 2025). We further honed the landscape variables by removing those that were insignificant (p > 0.05) via the Holm correction, and those that had high variance inflation factors (VIF > 10) as a metric of multicollinearity. The RDA reported the most important landscape variables for the natural enemies as a whole at each buffer zone. We also did the same thing for the herbivores/ non-natural enemies caught.
To look at individual taxa singularly, we ran a random-forest analysis one buffer zone at a time, using 500 runs the “Boruta” function in the “Boruta” package (Kursa and Rudnicki 2010) which selects the most important explanatory variable for each response. We checked for correlation between variables using the “ggcorrplot” function from the package “ggcorrplot” (Kassambara and Patil 2023). Then, we went on to individual GLMMs for each natural enemy taxon, using only the uncorrelated important landscape variables, and the random site nested in the farm and month nested in year, and proceeded to choose the best fit models using AIC and run estimated marginal means post-hoc tests. We ran additional models to see if there was an interaction between any of the landscape variables and MeSA.
4.2 Pollinators
4.2.1 Pollinators and MeSA
The response variables in the pollinator study were the pollinator counts. Pollinators had been identified at minimum to family, with some to species. For the analyses, we grouped species that were in the same genus together, therefore we can represent a diversity of insects that do not range in identification from family to species. Our pollinator taxa were either genera (such as Bombus spp.), monotypically represented genera (such as Apis mellifera L.), or family (such as Muscidae). We also further calculated Shannon diversity of pollinators at each site, as well as total abundance of pollinators at each site.
For the plant level, we began with Hellinger transforming the pollinator count data using “decostand” function, testing for dispersion homogeneity using the “betadisper” and “permutest” functions, then running a Permutational Multiple Analysis of Variance (PERMANOVA) to look at the effects of MeSA on the pollinator community with 999 permutations on a Bray-Curtis distances using “adonis2” function in RStudio (ver. 4.2; R Core Team 2024) all with the “vegan” package (Oksanen et al. 2025). After that, we looked at one pollinator taxa at a time and ran generalized linear mixed models (GLMMs) with MeSA with random factors of site nested in farm and month nested in year, using the “glmer” function in the “lme4” package in RStudio (Bates et al. 2015). When looking at one pollinator at a time we did not look at Hellinger transformed data, rather raw data. The distributions used for the GLMMs were either Poisson or Negative Binomial, and that was determined using the “overdisp_fun” function from the “PsychHelperFunctions” package (Huff and Papenmeier 2022). We also looked at the effects of MeSA on the total pollinator abundance and Shannon diversity of pollinators by running a Negative Binomial distributed GLMM and a Gamma log link shited by 0.1 GLMM respectively, both with the same random factors as earlier. Estimated marginal means post-hoc tests were run using the “emmeans” function in the “emmeans” package (Lenth 2025).
4.2.2 Pollinators and local management
Moving onto the field level, our explanatory variables were the number of sprays per pesticide type (insecticide, herbicide, and fungicides), cultivar, interactions, and again site nested in farm and month nested in year as random factors. We ran with the “glmer” function GLMMs for each pollinator taxon (Bates et al. 2015), and then chose best fit models by eliminating insignificant variables and looking for the lowest Akaike’s Information Criterion (AIC). By itself, the AIC value does not mean anything, but it is used in model comparison and selection (Symonds et al. 2011). Afterwards, we ran estimated marginal means post-hoc tests.
Since we also wanted to know how MeSA can affect the pollinator’s behavior towards the different management tactics, we ran another separate round of GLMMs. This time our explanatory variables were each of the field level variables interacting with MeSA, and the same random factors. We similarly chose the best fit models using AIC and ran estimated marginal means post-hoc tests.
4.2.3 Landscape level pollinators
Lastly for the landscape level, the explanatory variables were the all landscape factors per buffer zone (Shannon diversity of land covers, the total edge length between land covers, the largest patch index, numbers of patches of each land cover, and the percent area of each land cover), year, and the random site nested in farm. We ran a partial redundancy analysis (RDA) with the “rda” function in the “vegan” package (Oksanen et al. 2025), which is similar to a principal component analysis, but it considers a constrained environment, i.e the landscape variables. It was a partial RDA because we only wanted to account for the landscape while controlling the year and month, so we added those as covariates. For this analysis we looked at each buffer zone separately. Before running the partial RDA, we standardized the numeric explanatory variables to have a mean of 0 and standard deviation of 1, as well as the community response with the Hellinger transformation, both by utilizing the “decostand” function. We ran several rounds of RDA and performed forward selection on variables using the “ordiR2step” function from the “vegan” package (Oksanen et al. 2025). We further honed the landscape variables by removing those that were insignificant (p > 0.05) via the Holm correction, and those that had high variance inflation factors (VIF > 10) as a metric of multicollinearity. The RDA reported the most important landscape variables for the pollinators as a whole at each buffer zone.
To look at individual taxa singularly, we ran a random-forest analysis one buffer zone at a time, using 500 runs the “Boruta” function in the “Boruta” package (Kursa and Rudnicki 2010) which selects the most important explanatory variable for each response. We checked for correlation between variables using the “ggcorrplot” function from the package “ggcorrplot” (Kassambara and Patil 2023). Then, we went on to individual GLMMs for each pollinator taxon, using only the uncorrelated important landscape variables, and the random site nested in the farm and month nested in year, and proceeded to choose the best fit models using AIC and run estimated marginal means post-hoc tests. We ran additional models to see if there was an interaction between any of the landscape variables and MeSA.
4.3 Predation
In the predation study, the response variable was the proportion of eggs eaten. We did not count the three egg masses at each location as individual replicates, instead we calculated the predation rates from the combined egg loss at each site. Predation rates then underwent an arcsine square-root transformation before analyzing under 0.1 shifted Gamma log link distributed GLMMs. Initially, our fixed variable was just the presence or absence of MeSA. Then, we looked at the field level and whether the pesticide regime or cultivar affected predation. We additionally looked to see if there was an interaction between the management and MeSA. Lastly, we investigated the landscape variables by selecting the most important with the “Boruta” function, and then seeing if there was any influence. We then also looked to see if there was an interaction between the landscape factors and MeSA. The random factors were bed nested infarm and month nested in year. All GLMMs were followed by AIC model selection and then estimated marginal means post-hoc tests.
4.4 Pollination
We approached the statistics very similarly for the pollination study, except the response variables were now yields (berry count, average berry volume, and average berry weight per sample) as a proxy for pollination. On the plant level, our explanatory variable was MeSA, with the year and the site nested in the farm as random factors, and we ran G/LMMs for each yield measurement. Berry count was run with a Negative Binomial distribution, the volume had a normal distribution, and the mass was run as 0.1 shifted Gamma distribution with a log link. After that, we looked at the field level, using the pesticide regiment, cultivar, and the random effects; however, since these cultivars have been bred to differ in yields (Vorsa and Johnson-Cicalese 2012), we ran the cultivar and pesticide models separately. Subsequently, we looked at the landscape variables. We selected the most important landscape variables, we ran a random-forest analysis one buffer zone at a time with the “Boruta” function from the “Boruta” package (Kursa and Rudnicki 2010) followed by a correlation matrix using “ggcorrplot” from package “ggcorrplot” to ensure no cross-talk (Kassambara and Patil 2023). All the models were refined using the AIC values and had estimated marginal means post-hoc tests run.
Results
1. Natural enemies
1.1 Natural enemy and pest communities (sticky traps) and MeSA
We caught 418,173 natural enemy individuals across 28 taxa (mostly families), and 1,632,406 herbivorous individuals across 30 taxa. We found that both natural enemies (F = 8.4419, df = 1, p = 0.001) and herbivores (F = 7.0198, df = 1, PERMANOVA p = 0.002) were affected by MeSA as a whole, and that both natural enemies (F = 0.0005, df = 1, p = 0.986) and herbivores (F = 0.0753, df = 1, p = 0.782) had a homogenous dispersion by MeSA. The total abundance (χ² = 8.5727, df = 1, p = 0.0034) and Shannon diversity (χ² = 6.2833, df = 1, p = 0.0122) of natural enemies both increased with the presence of the MeSA lure, while the total herbivore count marginally increased (χ² = 3.7206, df = 1, p = 0.0537) and diversity marginally decreased (χ² = 2.8981, df = 1, p = 0.0887). Of our natural enemies, we specifically found Syrphidae, Dolichopodidae, Tachinidae, Tabanidae, Empididae, Coccinellidae, Anthocoridae, Pentatomidae, Araneae, and Neuroptera to be significantly attracted to MeSA, while Formicidae was significantly repelled and Phlaeothripidae was marginally repelled (Table 1).
1.2 Natural enemies and local management
Insecticides significantly impacted Tachinidae (χ² = 4.6236, df = 1, p = 0.0315), Anthocoridae (χ² = 9.1491, df = 1, p = 0.0025), Phlaeothripidae (χ² = 6.2833, df = 1, p = 0.0022), parasitoid wasps (χ² = 8.2869, df = 1, p = 0.004), spiders (χ² = 4.7398, df = 1, p = 0.0295), and marginally Tettigoniidae (χ² = 3.6945, df = 1, p = 0.0564). Interestingly, only Tachinidae and Tettigoniidae have a negative relationship with insecticides, while Anthocoridae, Phlaeothripidae, parasitoid wasps, and spiders all increase with increasing insecticides. Fungicides also had many positive relationships with natural enemies, specifically with Syrphidae (χ² = 9.1279, df = 1, p = 0.0025), Tachinidae (χ² = 13.876, df = 1, p = 0.0002), Cantharidae (χ² = 17.6398 df = 1, p < 0.0001), Lampyridae (χ² = 4.3945, df = 1, p = 0.0361), Phlaeothripidae (χ² = 4.1093, df = 1, p = 0.0426), and parasitoid wasps (χ² = 24.4068, df = 1, p < 0.0001), while negatively affecting Neuroptera (χ² = 13.8109, df = 1, p = 0.0002). As a whole, increasing fungicides led to an increasing total abundance of natural enemies (χ² = 10.5254, df = 1, p = 0.0012) but decreasing Shannon diversity of natural enemies (χ² = 5.0684, df = 1, p = 0.0244). Herbicides had a positive impact on Neuroptera (χ² = 4.3878, df = 1, p = 0.0362), and a negative impact on Staphylinidae (χ² = 3.9845, df = 1, p = 0.0459), parasitoid wasps (χ² = 4.066, df = 1, p = 0.0438), and marginally Formicidae (χ² = 3.6582, df = 1, p = 0.0558).
Cultivar was also significant more many natural enemies including Empididae (χ² = 10.7561, df = 4, p = 0.0294), Coccinellidae (χ² = 23.8127, df = 4, p < 0.0001), Lampyridae (χ² = 9.7443, df = 4, p = 0.045), Miridae (χ² = 10.023, df = 4, p = 0.04), Anthocoridae (χ² = 16.1984, df = 4, p = 0.0028), parasitoid wasps (χ² = 28.0678, df = 4, p < 0.0001), and to natural enemy abundance as a whole (χ² = 9.5996, df = 4, p = 0.0477). Specifically, Coccinellidae had higher counts in ‘Mullica Queen’ than ‘Stevens’, Miridae and parasitoids had higher counts in ‘Stevens’ than ‘Crimson Queen’, and there were also more parasitoid wasps in both ‘Ben Lear’ and ‘Stevens’ than ‘Mullica Queen’. Cultivar additionally had several significant interactions with the pesticide type for many natural enemies. Notably, increasing insecticides led to more Tabanidae catches in ‘Crimson Queen’ but less in ‘Early Black’ and ‘Mullica Queen’ (χ² = 9.9572, df = 4, p = 0.0412); more Empididae in ‘Early Black’ than ‘Mullica Queen’ (χ² = 15.0883, df = 4, p = 0.0045); more Coccinellidae in each of the cultivars except for less in ‘Mullica Queen’ (χ² = 10.4618, df = 4, p = 0.0333); more Anthocoridae in ‘Ben Lear’ but less in ‘Early Black’ (χ² = 11.636, df = 4, p = 0.0203); more Phlaeothripidae in ‘Stevens’ (χ² = 21.9715, df = 4, p = 0.0002); more spiders in ‘Crimson Queen’ (χ² = 13.426, df = 4, p = 0.0094); and overall lower Shannon diversity but specifically lower in ‘Stevens’ and ‘Early Black’ (χ² = 14.58577, df = 4, p = 0.0056). Increasing fungicides led to lower Dolichopodidae counts in all the cultivars except for ‘Stevens’ (χ² = 10.6639, df = 4, p = 0.0306); less Empididae in ‘Early Black’ than ‘Mullica Queen’ (χ² = 10.0291, df = 4, p = 0.0399); less Coccinellidae in ‘Mullica Queen’ and ‘Crimson Queen’ but more in the other cultivars (χ² = 103.322, df = 4, p = 0.0353); more Cantharidae in ‘Ben Lear’ (χ² = 38.3065, df = 4, p < 0.0001); more Cleridae in ‘Early Black’ (χ² = 11.8122, df = 4, p = 0.0188), more Anthocoridae in ‘Early Black’ than ‘Crimson Queen’ (χ² = 28.695, df = 4, p = 0.0003); less Formicidae in ‘Early Black’ than ‘Mullica Queen’ and ‘Crimson Queen’ (χ² = 9.4883, df = 4, p = 0.0499); more Phlaeothripidae in ‘Crimson Queen’ (χ² = 10.644, df = 4, p = 0.0309); more spiders in ‘Crimson Queen’ but less in ‘Mullica Queen’ (χ² = 20.5469, df = 4, p = 0.0004); more parasitoid wasps (χ² = 28.695, df = 4, p < 0.0001) and overall natural enemy abundance (χ² = 10.0925, df = 4, p = 0.0391) in ‘Ben Lear’, ‘Mullica Queen’, and ‘Stevens’ than the other two; and generally lower Shannon diversity of natural enemies especially in ‘Mullica Queen’ other than ‘Early Black’ which had higher diversity (χ² = 9.7712, df = 4, p = 0.0445). Lastly, increasing herbicides led to more Dolichopodidae especially in ‘Early Black’ then ‘Ben Lear’ (χ² = 17.9693, df = 4, p = 0.0013); and more Tachinidae especially in ‘Mullica Queen’ (χ² = 22.6794, df = 4, p = 0.0001).
The presence or absence of the MeSA bait also had several significant interactions with several of the management factors. For example, MeSA interacting with the number of fungicides significantly affected Syrphidae (χ² = 5.3252, df = 1, p = 0.021), Phoridae (χ² = 5.1641, df = 1, p = 0.0231), Cantharidae (χ² = 5.1152, df = 1, p = 0.0237), Cleridae (χ² = 7.8504, df = 1, p = 0.0051), Anthocoridae (χ² = 4.2981, df = 1, p = 0.0382), Formicidae (χ² = 8.7649, df = 1, p = 0.0031), Neuroptera (χ² = 8.7457, df = 1, p = 0.0031), and Megaloptera (χ² = 23.6866, df = 1, p < 0.0001). Syrphidae had a more positive relationship with fungicides in the baited traps, meanwhile Cantharidae and Cleridae had more positive relationships in the unbaited traps. Formicidae had a more negative fungicide relationship in the unbaited traps, and Neuroptera and Megaloptera had more negative relationships with the baited traps. Insecticides had an interaction in Odonata (χ² = 4.6427, df = 1, p = 0.0312) and marginally in parasitoid wasps (χ² = 3.6308, df = 1, p = 0.0567) in which both insect groups did better with increasing insecticides when a MeSA bait was present. Similarly, increasing herbicides increased Dolichopodidae (χ² = 5.1802, df = 1, p = 0.0228) in the presence of MeSA. Certain cultivars were more attractive to specific natural enemies with MeSA, such as baited ‘Stevens’, ‘Early Black’, and ‘Ben Lear’ attracting more Empididae (χ² = 19.6767, df = 4, p = 0.0006) and baited ‘Mullica Queen’ and ‘Crimson Queen’ attracting more Neuroptera (χ² = 9.8708, df = 4, p = 0.0426) than those same cultivars unbaited. Unbaited ‘Early Black’ however attracted more Cleridae (χ² = 15.5297, df = 4, p = 0.0037) than baited ‘Early Black’.
3.3 Natural enemy and herbivore communities on the landscape scale
Upon investigating the largest buffer spatial scale (1500 m radius), we found that the landscape accounts for around 9.04% of the variation of natural enemy community composition across sites, while the month and year explain 20.46% of this variation, leaving 70.5% unexplained. Our model was significant (p = 0.001) and has an adjusted R2 that explains 8.54% of the variation. More specifically, the significant landscape variables were the percent of the zone that was agriculture (F = 51.3669, df = 1, p = 0.001), the percent zone wetland (F = 20.018, df = 1, p = 0.001), the largest patch index (F = 9.7766, df = 1, p = 0.001), the percent zone urban (F = 7.7503, df = 1, p = 0.001), the land cover diversity (F = 5.1008, df = 1, p = 0.002), and the number of forest patches (F = 7.1067, df = 1, p = 0.001) (Table 2). The first three canonical RDA axes were significant (p < 0.05). While looking at the this buffer zone for the herbivore communities, we found that the landscape explained 3.33% of variation, the month and year explained 32.35%, and 64.32% was unconstrained. Our model was significant (p = 0.001) and the adjusted R2 was 3.18%. The significant landscape variables for herbivores were the land cover diversity (F = 32.2278, df = 1, p = 0.001) and the percent of the zone that was barren (F = 8.7898, df = 1, p = 0.001)(Table 2). Both of the canonical RDA axes were significant (p < 0.05).
At the 500 m radius zone for natural enemies, the landscape explained 5.95%, the month and year explained 20.46% again, and now around 73.58% remained unexplained. The model was still significant (p = 0.001), with an adjusted R2 of 5.23%. This time the important variables were the number of agricultural patches (F = 13.6211, df = 1, p = 0.001), the number of waterbody patches (F = 9.3917, df = 1, p = 0.001), the percent zone agriculture (F = 12.3975, df = 1, p = 0.001), the number of urban patches (F = 7.6611, df = 1, p = 0.001), land cover diversity (F = 8.2044, df = 1, p = 0.001), number of wetland patches (F = 6.4299, df = 1, p = 0.001), and the class of the largest patch (F = 2.9743, df = 2, p = 0.004) (Table 2). The first two canonical RDA axes were significant (p < 0.05). For herbivores at this spatial scale, the landscape explained 2.21%, the month and year explained 32.35%, and 65.44% was unexplained. The model was significant (p = 0.001), and had an adjusted R2 of 1.97%. The important landscape variables were the numbers of of agricultural patches (F = 8.2495, df = 1, p = 0.001), number of waterbody patches (F = 5.8185, df = 1, p = 0.005), and total edge length (F = 12.6530, df = 1, p = 0.001) (Table 2). The first two canonical RDA axes significant again (p < 0.05).
Landscapes explained 5.15% of the variation of natural enemy community composition at the 250 m radius spatial scale, with 20.46% still explained by month and year, and 74.39% unexplained. The model had an adjusted R2 value of 4.61% and was significant (p = 0.001). At this scale, most important landscape variables were the class of the largest patch (F = 6.9662, df = 2, p = 0.001), the number of wetland patches (F = 13.6211, df = 1, p = 0.001), the number of waterbody patches (F = 17.8368, df = 1, p = 0.001), the total edge length (F = 7.1098, df = 1, p = 0.001), and the percent zone made up of water (F = 3.9148, df = 1, p = 0.005) (Table 2). The first three canonical RDA axes were significant (p < 0.05). At this scale for the herbivores, landscape accounted for 3.73% of variation, month and year accounted for 32.35% of variation, leaving 63.92% unconstrained. That model was also significant (p = 0.001) with an R2 value of 3.43%. The number of wetland patches (F = 15.9674, df = 1, p = 0.001), number of urban patches (F = 14.9268, df = 1, p = 0.001), total edge length (F = 7.9181, df = 1, p = 0.002), and number of forest patches (F = 7.3662, df = 1, p = 0.001) were significant variables (Table 2), as well as the first two canonical RDA axes (p < 0.05).
The smallest buffer zone (100 m radius) had only 3.1% of the variation of natural enemy community explained by the landscape, with 20.46% still explained by month and year, and 76.44% unexplained. Our model was still significant (p = 0.001), and the adjusted R2 value was 2.73%. The important landscape variables at this spatial scale were percent zone water (F = 9.8112, df = 1, p = 0.001), percent zone urban (F = 7.8043, df = 1, p = 0.001), number of waterbody patches (F = 7.5933, df = 1, p = 0.001), and the land cover diversity (F = 6.8724, df = 1, p = 0.001) (Table 2). Only the first canonical RDA axis was significant (p < 0.05). For herbivores, only around 1.59% of variation was explained by the landscape, 32.35% by month and year, and around 66.07% unexplained. This model was significant (p = 0.001) and had an adjusted R2 value of of 1.43%. The landscape variables that were important were the percent zone agriculture (F = 11.6092, df = 1, p = 0.001) and numbers of urban patches (F = 7.4104, df = 1, p = 0.003) (Table 2), and both the canonical RDA axes were significant (p < 0.05).
The individual results of the effects of the landscape per taxon are still underway.
2. Pollinators
2.1 Pollinator community (pan traps) and MeSA
We caught 121,000 individual arthropods over the course of this study, 11,145 of which we believe could be potential pollinators. We found that as a whole, pollinators were impacted by MeSA (F = 2.6819, df = 1, p = 0.011) and had homogeneous dispersions with and without MeSA (F = 2.3794, df = 1, p = 0.134). The total abundance of pollinators caught were higher in the MeSA baited traps than the unbaited ones (χ² = 23.196, df = 1, p < 0.0001), however there was not difference in Shannon diversity (χ² = 1.1504, df = 1, p = 0.2835). Upon looking at each individual pollinator taxa, MeSA significantly attracted A. mellifera (χ² = 22.913, df = 1, p < 0.0001), Cerceris spp. (χ² = 8.946, df = 1, p = 0.0028), and Toxomerus marginatus Say (χ² = 27.077, df = 1, p < 0.0001), and marginally attracted Tiphiidae (χ² = 3.6009, df = 1, p = 0.0578). However, MeSA repelled Agapostemon spp. (χ² = 3.8441, df = 1, p = 0.0499), Megachile spp. (χ² = 5.0084, df = 1, p = 0.0252), Lampyridae (χ² = 5.2849, df = 1, p = 0.0215), and marginally repelled Vespidae (χ² = 3.4778, df = 1, p = 0.0622) (Table 1).
2.2 Pollinators and local management
Insecticides, fungicides, and herbicides all had different effects on different pollinators. Increasing insecticides positively impacted many Dipteran pollinators, such as T. marginatus (χ² = 10.3423, df = 1, p = 0.0013), Eupeodes americanus Wiedemann (χ² = 8.8751, df = 1, p = 0.0029), Chalcosyrphus sp. (χ² = 3.9819, df = 1, p = 0.0460), Dolichopodidae (χ² = 13.543, df = 1, p = 0.0002), Muscidae (χ² = 12.4308, df = 1, p = 0.0004), in addition to Agapostemon splendens Lepeletier (Hymenoptera: Halictidae) (χ² = 4.6665, df = 1, p = 0.0308) and marginally Mordellidae (χ² = 3.5889, df = 1, p = 0.0582). Interestingly, increasing insecticides also led to an increase in the Shannon diversity of pollinators (χ² = 10.259, df = 1, p = 0.0014). However, increasing insecticides negatively impacted Megachile spp. (χ² = 3.9453, df = 1, p = 0.0470), Bombus spp. (χ² = 4.3881, df = 1, p = 0.0362), and Cerceris spp. (χ² = 7.1459, df = 1, p = 0.0075). Fungicides negatively affected A. mellifera (χ² = 8.6938, df = 1, p = 0.0032), Ammophila spp. (χ² = 5.6127, df = 1, p = 0.0178), and Dolichopodidae (χ² = 3.922, df = 1, p = 0.0477), as well as marginally negatively impacting Xylocopa virginica L. (Hymenoptera: Apidae) (χ² = 3.7352, df = 1, p = 0.0533). Fungicides positively impacted Tachinidae (χ² = 5.1534, df = 1, p = 0.0232), and marginally Chrysididae (χ² = 3.5961, df = 1, p = 0.0579) and Mordellidae (χ² = 3.539, df = 1, p = 0.0599). The total pollinator abundance decreased with fungicides (χ² = 4.5058, df = 1, p = 0.0338). Interestingly, herbicides positively influenced pollinators, specifically Ammophila spp. (χ² = 7.1757, df = 1, p = 0.0074) and Tiphiidae (χ² = 6.9843, df = 1, p = 0.0082).
Cultivar was also an important factor for several pollinators including A. mellifera (χ² = 10.8963, df = 4, p = 0.0278), Bombus spp. (χ² = 15.8457, df = 4, p = 0.0032), T. marginatus (χ² = 21.7845, df = 4, p = 0.0002), Dielis plumipes Drury (Hymenoptera: Scoliidae) (χ² = 12.5282, df = 4, p = 0.0138), Tiphiidae (χ² = 12.1543, df = 4, p = 0.0162), and Tachinidae (χ² = 20.4661, df = 4, p = 0.0004), as well as for the total abundance of pollinators (χ² = 10.1531, df = 4, p = 0.0379), where there were more A. mellifera and pollinators in total in ‘Mullica Queen’ than ‘Ben Lear’, and specifically more Tachinidae in each of ‘Crimson Queen’, ‘Early Black’, ‘Stevens’ than ‘Ben Lear’. There were also strong interactions between fungicides and cultivars for many pollinators such as Bombus spp. (χ² = 16.0895, df = 4, p = 0.0029), T. marginatus (χ² = 23.6143, df = 4, p < 0.0001), D. plumipes (χ² = 12.9696, df = 4, p = 0.0114), and marginally for Megachile spp. (χ² = 9.2815, df = 4, p = 0.0544). In these interactions, increasing fungicides led to lower Bombus spp. catch most notably in ‘Early Black’ then ‘Crimson Queen’; overall less T. marginatus particularly in ‘Mullica Queen’, with an exception of higher counts in ‘Crimson Queen’; as well as less D. plumipes in ‘Early Black’ and ‘Stevens’ but more in ‘Crimson Queen’. Additionally, there were interactions between cultivar and insecticides in which more insecticides led to more Muscidae in ‘Stevens’ (χ² = 18.3105, df = 4, p = 0.0011); and more Tachinidae in ‘Crimson Queen’ and ‘Stevens’ (χ² = 13.4705, df = 4, p = 0.0092). When herbicides increased there was also an increase in T. marginatus in ‘Crimson Queen’ and ‘Mullica Queen’, but a decrease in ‘Ben Lear’ (χ² = 8.666, df = 4, p = 0.0341).
The MeSA bait also had some interactions with the local management. There were higher catches of Lasioglossum spp. with increasing insecticides in the unbaited traps (χ² = 4.5683, df = 1, p = 0.0326); increasing fungicides had lower A. mellifera counts in the unbaited traps (χ² = 4.1986, df = 1, p = 0.0406); and increasing herbicides had more Bombus spp. in the baited traps but less in the unbaited traps (χ² = 5.7119, df = 1, p = 0.0169). In terms of a MeSA by cultivar interaction, baited traps in beds of ‘Mullica Queen’ had more Lasioglossum spp. than the unbaited traps of the same cultivar, however unbaited traps in ‘Early Black’ beds caught more of that genus than the baited in the same beds (χ² = 13.6696, df = 4, p = 0.0084). There were also higher Bombus spp. counts in the unbaited ‘Crimson Queen’ traps than the baited ones (χ² = 9.7553, df = 4, p = 0.0448). Baited traps in beds of ‘Ben Lear’, ‘Crimson Queen’, and ‘Early Black’ all had higher catches of T. marginatus than the corresponding unbaited traps (χ² = 10.7532, df = 4, p = 0.0295).
2.3 Pollinators and landscape
At the 1500 m radius buffer zone, 6.82% of the variation of pollinator community composition across the sites were explained by the landscape, with 6.7% explained by the month and year, and 86.48% unexplained. This partial RDA model was significant (p = 0.001) and there was an adjusted R2 value of 6.41%. To explain the variation by landscape, the most important variables were the land cover diversity (F = 26.6449, df = 1, p = 0.001) and the total edge length (F = 4.4922, df = 1, p = 0.003) (Table 2). The first canonical RDA axis was also significant (p < 0.05).
For the 500 m radius zone, landscape accounts for 3.75% of variation, month and year for 6.7% of variation, and an unconstrained 89.55%. The model was significant (p = 0.001) and had an adjusted R2 value of 3.32%. The most important landscape variables were the percent of zone forest (F = 11.5487, df = 1, p = 0.001) and the percent zone wetlands (F = 5.0009, df = 1, p = 0.001) (Table 2), and only the first canonical RDA axis was significant (p < 0.05).
When looking at the 250 m radius spatial scale, landscape explained 3.71% of variation, month and year for 6.7% of variation, and 89.59% unexplained. The model was significant (p = 0.001), and the adjusted R2 value was 3.05% with the number of wetland patches (F = 7.5286, df = 1, p = 0.001), number of waterbody patches (F = 5.2959, df = 1, p = 0.001), and the percent zone forest (F = 3.5111, df = 1, p = 0.001) being the most important (Table 2). The first canonical RDA axis was significant (p < 0.05).
Lastly, in the 100 m radius zone, 1.72% of variation was explained by the landscape, 6.7% explained by month and year, and 91.58% unexplained. The model was significant (p = 0.001), and the adjusted R2 value was 1.49%. The most important landscape variable was the number of forest patches (F = 7.4161, df = 1, p = 0.001) (Table 2), and the canonical RDA axis was significant too (p < 0.05).
Upon looking at individual taxa, we found several trends (Table 3). We first off looked at A. mellifera, since it was the dominant species, and we found that it had a negative relationship with forest at the two smaller spatial scales, and a negative relationship with the total edge length at the larger spatial scales. The only significantly positive relation it had was at the 1500 m zone, with increasing water as well as increasing urbanized land, there were more A. mellifera. These all make sense, because the honeybees are manually brought in by the growers and placed consistently around the farm. That would mean that there are more A. mellifera in areas farther from forest and closer to roads. This interestingly portrayed itself when looking at the Shannon diversity and total abundances of pollinators. The Shannon diversity of pollinators had a positive relation with forest at the lower scales, total abundance had a negative relation at the smallest scale with percent zone forest, and increasing total edge length had lower total abundance at the larger scales. The Shannon diversity of pollinators was negatively affected by increasing agriculture at the larger spatial scales. There was a discrepancy between the 500 m and 1500 m zones for the Shannon diversity of pollinators, where the first one had a positive relationship with the land cover diversity and the latter a negative (Table 3).
Our next most commonly caught single species pollinator, T. marginatus, barely had any significant landscape influence, with only a decreasing amount with increasing urban patches in the largest spatial scale (Table 3). Other numerous pollinators included Muscidae, Dolichopodidae, Tachinidae, D. plumipes, Bombus spp., and Lasioglossum spp. Muscids tended to dislike higher water and urban patches at larger scales, but enjoy a diverse landscape at the lowest scales. Dolichopodids also were less frequent with increasing water and barren land at the large spatial scales. Similarly, Tachinidae decreased with increasing wetlands, increasing agriculture, and increasing urban land in the largest buffer zone, but increased with land cover diversity at the 500 m zone and total edge length at the 250 m zone. Dielis plumipes had a positive relationship with forest and negative relationship with water at the largest zone, and an overall negative relationship with agriculture. It also increased with land cover diversity on the smallest scale but decreased with land cover diversity on the largest scale. Bombus spp. decreased with water, agriculture, and urban lands, but increased with wetlands at the largest scale and land cover diversity at the smaller scales. Lastly, except for the smallest scale where it was positive, Lasioglossum spp. tended to stay away from areas with more water, in addition to being less present with areas of higher agriculture. However, this genus did have a positive relation with forest and urban lands at larger spatial scales and edge length and wetlands at smaller scales (Table 3).
We have not yet analyzed the interactions between the landscape and MeSA.
3. Predation
The predation study is currently being analyzed so the results are not complete.
4. Pollination
4.1 Pollination services and MeSA
We found that the cages baited with MeSA resulted in higher berry count than those without the MeSA (χ² = 4.4291, df = 1, p = 0.03533). Baited cages had on average 67.18 ± 5.08 cranberries, while unbaited cages had 55.78 ± 4.37. However, MeSA did not have any influence on either volume or mass.
4.2 Pollination services and local management
The pesticides that significantly affected berry count were fungicides (χ² = 19.3693, df = 1, p < 0.0001) and herbicides (χ² = 5.3155, df = 1, p = 0.02114), of which both led to higher amounts of berries. With increasing insecticides, there was increasing volume (χ² = 10.0764, df = 1, p = 0.0015) and mass (χ² = 10.2377, df = 1, p = 0.0014) of the berries. Similarly, increasing herbicides also resulted in bigger (χ² = 16.393, df = 1, p < 0.0001) and heavier (χ² = 4.5407, df = 1, p = 0.0331) fruit.
Cultivar very clearly impacted berry count (χ² = 27.5, df = 4, p < 0.0001), average volume (χ² = 178.19, df = 4, p < 0.0001), and average mass (χ² = 135.31, df = 4, p < 0.0001). ‘Mullica Queen’ had more berries than either ‘Ben Lear’ or ‘Crimson Queen’, and ‘Stevens’ also had more than ‘Crimson Queen’. ‘Crimson Queen’ had the most volume and mass, being larger than every other cultivar other than ‘Mullica Queen’. ‘Mullica Queen’ was statistically similar to ‘Stevens’ in size and weight, and both were heavier than ‘Ben Lear’ although ‘Stevens’ was comparable in size to ‘Ben Lear’. In turn, ‘Ben Lear’ was both bigger and heavier than ‘Early Black’.
Only insecticides significantly interacted with MeSA to affect berry count (χ² = 13.5069, df = 1, p = 0.0002), in which increasing insecticides led to more berries in the baited cages. Methyl salicylate marginally interacted with cultivar to impact berry count (χ² = 9.3328, df = 4, p = 0.0533) and significantly interacted to impact berry size (χ² = 9.8437, df = 4, p = 0.0431), where MeSA led to increased berry count and size specifically in ‘Crimson Queen’.
4.3 Pollination and landscape
Pollination and landscape is still being analyzed and is not yet complete.
Methyl salicylate (MeSA) can be used to attract many beneficial insects. Previous studies have shown that many natural enemies, notably Syrphidae, Coccinellidae, and Neuroptera, will come to a MeSA lure (Rodriguez-Saona et al. 2012). Our study corroborated this, showing increased levels of several natural enemy families at the MeSA baited traps. Syrphidae, whose larvae are predaceous, also are active pollinators, which doubles their ecosystem services. We found that MeSA also attracts honeybees, Apis mellifera, which has never been shown before. In the context of applied agriculture that relies on pollinators, discovering a potential field tested bee attractant can be monumental in increasing yields. We found that MeSA was able to increase pollination services, as seen by the heightened fruit count.
This study also aimed to see the relationships that local management, such as insecticides, fungicides, herbicides, and cultivar, have on the beneficial insect community, and those further interactions with MeSA. We found that all these factors are unique to the beneficial insect taxon. One interesting trend in the pollinator/pollination study we found was that specifically the variety Crimson Queen had more and bigger fruits when exposed to the MeSA lure. At the same time, when faced with increasing fungicides and herbicides, the numbers of the most commonly found syrphid, Toxomerus marginatus, increased in that variety, especially when the lure was present. The presence and attraction of syrphids to MeSA may increase both pollination and predation at sites were the fly can be located. From the pollinator project, syrphids tended to have a positive relationship with insecticides, and from the natural enemy study they had a positive relationship with fungicides especially under the influence of MeSA. These are both interesting results and are good for farmers that may be worried about killing their syrphids. There are no Dipteran severe pests of cranberries (other than occasional gall midge), and with increased restrictions of pesticides it seems the growers do not have all too many broader spectrum compounds that are lethal to that group.
The largest spatial scales (1500 m radius) are seemingly are the best explaining the variation across the communities of natural enemies, pollinators, as well as herbivores. Most notably, the amount of water and agriculture followed by wetlands, were trends of important land covers amongst all the guilds and several spatial scales. At both the 500 m and 1500 m spatial scales, increasing agriculture decreased the Shannon diversity of pollinators, which may imply that other land covers may be critical in maintaining pollinators (particularly non-Apis mellifera which almost definitely drove the results). Interestingly, at the 500 m scale, increasing land cover diversity also increases pollinator diversity, however this had the opposite effect at the 1500 m scale. This may indicate that A. mellifera are utilizing areas larger than 500 m to forage non-crop land. That being said, A. mellifera tended to not like higher edge lengths at larger spatial scales nor increasing forest at smaller ones. There were other common pollinators that preferred forests at large spatial scales, like D. plumipes, Lasioglossum spp, Megachile spp, and Augochlorella spp. All in all different pollinators had preferences towards different landscapes compositions and configurations at different spatial scales. We have not yet investigated the interaction between landscapes and MeSA, nor have we investigated yet the landscape scale for natural enemies, predation, or pollination.
Citations
Adamczyk, J., Tiede, D., 2017. ZonalMetrics - a Python toolbox for zonal landscape structure analysis. Computers & Geosciences 99, 91–99. doi: 10.1016/j.cageo.2016.11.005
Bates D, Mächler M, Bolker B, Walker S (2015). “Fitting Linear Mixed-Effects Models Using lme4.” Journal of Statistical Software, 67(1), 1–48. doi:10.18637/jss.v067.i01
Ben-Zvi, Y., Rodriguez-Saona, C. 2023. Major insect pests of cranberries in New Jersey. New Jersey Agricultural Experiment Station, Cooperative Extension Fact Sheet FS1354. https://njaes.rutgers.edu/fs1354/
Huff M, Papenmeier F (2022). PsychHelperFunctions. R package version 0.3.9, https://github.com/markushuff/PsychHelperFunctions/
Kassambara A, Patil I (2023). Package ‘ggcorrplot’: Visualization of a Correlation Matrix using 'ggplot2'. R package version 0.1.4 https://github.com/kassambara/ggcorrplot
Kursa, M. B., & Rudnicki, W. R. (2010). Feature Selection with the Boruta Package. Journal of Statistical Software, 36(11), 1–13. https://doi.org/10.18637/jss.v036.i11
Lenth R (2025). emmeans: Estimated Marginal Means, aka Least-Squares Means. R package version 1.10.6-090003, doi: 10.32614/CRAN.package.emmeans
Oksanen J, Simpson G, Blanchet F, Kindt R, Legendre P, Minchin P, O'Hara R, Solymos P, Stevens M, Szoecs E, Wagner H, Barbour M, Bedward M, Bolker B, Borcard D, Borman T, Carvalho G, Chirico M, De Caceres M, Durand S, Evangelista H, FitzJohn R, Friendly M, Furneaux B, Hannigan G, Hill M, Lahti L, McGlinn D, Ouellette M, Ribeiro Cunha E, Smith T, Stier A, Ter Braak C, Weedon J (2025). vegan: Community Ecology Package. R package version 2.7-0, https://github.com/vegandevs/vegan, https://vegandevs.github.io/vegan/.
Rodriguez-Saona, C., Byers, J.A., Schiffhauer, D. (2012). Effect of trap color and height on captures of blunt-nosed and sharp-nosed leafhoppers (Hemiptera: Cicadellidae) and non-target arthropods in cranberry bogs. Crop Protection, 40, pp.132-144. https://doi.org/10.1016/j.cropro.2012.05.005
Rodriguez-Saona, C., Kaplan, I., Braasch, J., Chinnasamy, D., Williams, L. (2011). Field responses of predaceous arthropods to methyl salicylate: A meta-analysis and case study in cranberries. Biological Control, 59(2), pp.294-303. https://doi.org/10.1016/j.biocontrol.2011.06.017
Rodriguez-Saona, C., Urbaneja-Bernat, P., Salamanca, J., Garzon-Tovar, V. (2020). Interactive Effects of an Herbivore-Induced Plant Volatile and Color on an Insect Community in Cranberry. Insects 11(8), pp.524. https://doi.org/10.3390/insects11080524
Symonds, M. R., E., Moussalli, A. (2011). A brief guide to model selection, multimodel inference and model averaging in behavioural ecology using Akaike's information criterion. Behavioral Ecology and Sociobiology, 65(1), 13-21. https://doi.org/10.1007/s00265-010-1037-6
Education & Outreach Activities and Participation Summary
Participation Summary:
The ideas in this proposed research are novel and applicable to cranberry growers in New Jersey and the USA, my target audience. Since all the research will be done at commercial farms, there will be an opportunity for regular communication with growers. All the farms on which the research will be conducted have a long history in New Jersey, with the youngest one being over a century old and one of them being amongst the biggest cranberry producers in the world. Thus, this research will likely have profound impact on the whole cranberry industry in New Jersey and the USA.
By working directly with growers, I will be providing them with information on my research progress. Information on biological control and pollination services in cranberries will be provided to growers through various media sources. Factsheets will be written on how to identify and conserve natural enemies of herbivores and pollinators. Another factsheet was already written on New Jersey cranberry pests. This information will be given to growers also through bogs published in the Rutgers Plant & Pest Advisory.
Another important method of outreach is by attending and presenting at grower meetings and scientific conferences. I presented findings from this research at the American Cranberry Growers Association winter meeting, the North American Cranberry Research and Extension Workers conference, and at the Entomological Society of America branch and national meetings. I also plan on writing scientific articles including a meta-analysis about how and what synthetic herbivore-induced plant volatiles are used to attract natural enemies of cultivated crops as well as a synthesis literature review of the past thirty years of integrated pest management practices against insect pests of cranberries (which was published in 2023). These publications will be directed at both scientists and farmers, who can hopefully learn how to utilize plant volatiles and see the history of cranberry pest management.
2023 Update
I wrote and published a literature review on cranberry pest management from the past thirty years as well as a factsheet on cranberry pests of New Jersey, which was distributed at the American Cranberry Growers Association in January 2024. I also attended and presented at the North American Cranberry Research and Extension Workers meeting in August 2023 and the International Society of Chemical Ecology meeting in July 2023.
Advances in cranberry insect pest management: A literature synthesis
Major Insect Pests of Cranberries in New Jersey
2024-now Update
In 2024, I participated in a two-week short-course focused on chemical ecology, and also presented my work to other students and professors that work in similar fields as I do. I also presented my work at both the Entomological Society of America branch and national meetings in Morgantown, West Virginia and Phoenix, Arizona respectively. Additionally, I published a paper on the behaviors and performance of a parasitoid wasp in differentiating between two cranberry pests (Ben-Zvi and Rodriguez-Saona 2024). I gave a presentation to the farmers in the American Cranberry Growers Association winter meeting in 2025. Lastly, I took part in creating a series of 3 educational and informational videos on a severe cranberry pest. These videos will go on to teach growers about the pest, how to monitor for it, and what steps we are taking to mitigate it. These videos are not yet officially published yet.
Project Outcomes
My research will theoretically help farmers be more sustainable. Results from my research will suggest to farmers methods to consider their landscape when managing their land, allowing for greater conservation of beneficial insects. Furthermore, since I am also looking at ecosystem services, I was be able to inform the farmers on how much the beneficial insects are actually providing. Additionally, although it was not the focus of my study, by analyzing the arthropod communities around cranberry farms, I was also able to get a read on the presence and locations of certain pests. In the winter 2025 American Cranberry Growers Association meeting, I presented a portion of my work and a lot of farmers came up to me afterwards to discuss my findings and how they can better sample certain arthropods in certain places, and they had some ideas of continuation projects based off my initial findings.
During the course of this project so far I learned how to talk and work with farmers and how to use softwares like ArcGISPro and Rstudio. These two skills helped me realize how unique landscape ecology is and how important it is to approach problems from a large scale, since surrounding habitats may have some strong impacts on sustainability and integrated pest and pollinator management. I have also strengthened my insect identification skills, and have by far learned the most about statistics. This project has allowed me to see certain future careers that can focus on applied chemical ecology and landscape ecology.
There were several challenging aspects of this project. On the physical side, this took a lot of time staring through microscopes particularly at bright yellow sticky traps. I spent my time trying to count everything, but if I just focused on one group of arthropods, it would have been easily (although I would have gotten less information out). That being said, there is no one sampling method that can cover all arthropods, so my yellow sticky traps and white pan traps were all I was able to manage.
Additionally, figuring out which statistics to use when you have many multiple explanatory potential variables can get pretty confusing. I have spent a lot of time researching on what are the best statistical methods for this project, and I honestly still do not know 100% if I overdid, underdid, or just did it correctly. I thoroughly explained every step of my statistical methods here in hopes that future projects in similar fields can get some direction in the matter.
Lastly, some future directions of this project include more in depth analyses. For instance, I look at the impact of insecticides, but I don't look at the type or the rates, simply the number. Another future direction would to be incorporate other factors, such as prevalence of historic forest fire data, since fires are relatively common in the region, and that can be a major factor that impacts insects. Another future direction would be to specifically analyze the pest complex in cranberries, although that may require more targeted sampling methods. That information can further be used by farmers to know which areas of their farms are more at risk by certain pests, therefore be able to spot treat or use preventative methods at high risk locations.