Caneberries are trellised to facilitate harvest and agrochemical applications as well as to improve crop yield and quality. Trellising can also increase airflow and light penetration within the canopy and affect its microclimate. We compared an experimental trellis that split the canopy into halves to standard I- and V-trellises, measuring Drosophila suzukii (Matsumura) fruit infestation as well as canopy temperature and relative humidity in raspberries at two commercial you-pick diversified farms. To evaluate the combined effects of trellising systems and pruning, we pruned one half of each row in blackberry plantings at two research farms and assessed D. suzukii infestation, canopy microclimate (temperature, relative humidity, and light intensity), fruit quality parameters (interior temperature, total soluble solids, and penetration force), and spray coverage/deposition. Trellis installation costs, labor inputs, and yield were used to further evaluate the trellis systems from an economic perspective. Fruit quality was not affected by trellising or pruning and lower total yield was observed in the experimental trellis treatment on one farm. Although D. suzukii infestation was only affected by trellising and pruning at one site, we observed a relationship between higher temperatures and reduced infestation on nearly all farms. Occasionally, lower relative humidity and high light intensity corresponded with lower infestation. Ultimately, the experimental trellis was less economically efficient than other trellising systems and our ability to successfully manipulate habitat favorability varied in a site-specific manner. Drosophila suzukii management approaches that rely upon unfavorable conditions are likely to be more effective in hot, dry regions.

Sites and experimental design

In 2019 and 2020, an experimental trellis that split the canopy into halves was compared with a growers’ standard I-trellis (Site 1) or T-trellis (Site 2) at two commercial fruit farms in central Maryland in fall-bearing red raspberries (Rubus idaeus L., Site 1 ‘Nantahala’, Site 2 ‘Caroline’). Plantings were maintained following standard commercial practices for primocane fruiting raspberries, and in both years, growers applied insecticides as needed. Three treatment plots per trellising system were compared at each site. At Site 1 rows were spaced 4.0m apart, with treatment plots between 6.5m to 9m in length distributed across three rows. At Site 2 rows were spaced 3.4m apart, with treatment plots between 9.5m to 10.5m in length distributed across three rows. At both sites, up to 20 apparently marketable fruit (visually undamaged and texturally relatively firm) were collected from each row (10 from each side of the row) to evaluate D. suzukii infestation. Due to fruit availability, mean fruit sample size per plot across sites and years was 17.82 ± 0.23 with a range from 2 to 20. Fruit were collected weekly for a period of 10 weeks from July to September in 2019 and for 7 weeks from August to September in 2020. Data loggers were installed in all replicate plots to collect temperature and relative humidity data during the growing period. Microclimate data and D. suzukii infestation were the only metrics measured on grower cooperators’ farms.

In 2019 and 2020, the combined effects of trellising systems and pruning were examined in 4-year old ‘Prime Ark^® 45’ primocane blackberries (Rubus L. subgenus Rubus Watson) at the Wye Research and Education Center (WYE; Queen Anne’s County, MD) and at the Western Maryland Research and Education Center (WMREC; Washington County, MD). Plantings were maintained with standard herbicide applications and dormant pruning practices; however, no insecticides or fungicides were applied at WYE in either year of the study. At WMREC, lime-sulfur was applied in March of both years and no other insecticide or fungicide applications were made. An experimental V-trellis that split canopies into halves was installed in two rows of the planting while two rows maintained the previously installed standard trellis (WYE: V-trellis; WMREC: I-trellis), with trellis treatments arranged such that they alternated by row. In order to keep the growing canes within the trellis, canes needed to be rearranged and attached to the wire during the middle of the growing season (harvest week 4 – 6). Rows were spaced 3.0m apart at both sites. Treatment plot length differed at each farm with 10.5m plots at WYE and 13.4m at WMREC. In addition to trellising, pruning was conducted in July of both years by first removing all canes less than 6 mm in diameter in all plots. Then in one half of each trellis plot (WYE 5.25m; WMREC: 6.7m) the remaining canes per meter were counted and the number of canes that would represent a 25% reduction in canes was calculated and removed. In 2019 9 canes per meter (6 from the outer and 3 from the inner part of the plant) and in 2020 3 canes per meter (2 from the outer and 1 from the inner part of the plant) were removed at WYE. In 2019 and 2020 6 canes per meter (3 from the outer and 3 from the inner part of the plant) were removed at WMREC. Outer canes were removed from both sides of the row in an alternating pattern. After pruning, data loggers were installed in all replicates to measure temperature, relative humidity and light intensity during the growing season. Fruit temperature measurements were recorded in the field. All ripe fruit on both sites of the row were harvested weekly and collected in separate containers for each replicate plot. Harvested fruit were chilled during transportation from the farm to the laboratory.

In the laboratory, fruit were assessed for damage, sorted into marketable fruit and unmarketable fruit, and weighed. Fruit were considered unmarketable due to D. suzukii damage (visible larvae or soft leaky fruit), other insect feeding, evidence of pathogens, or other visible physical damage. Yield was calculated as grams per row meter (total length of both sides of the row). As available, fruit from each category and treatment replicate were randomly selected for D. suzukii infestation and quality measurements. Infestation measurements were prioritized and up to 10 marketable and 10 unmarketable fruit from each replicate were evaluated separately. Because fruit availability was occasionally low, mean fruit sample size per treatment replicate across marketability, research farms and years was 9.65 ± 0.05 and ranged from 1 to 10 fruit, with 2 samples missing.

Drosophila suzukii infestation

Infestation levels were evaluated weekly throughout the fruiting season at all sites. Larval extraction methods similar to Van Timmeren et al. (2017) were used to measure infestation. Briefly, fruit were weighed before larval extraction to calculate D. suzukii (g fruit)^-1. Fruit were gently crushed to break open tissues and soaked in a 50 g L^-1 table sugar water solution for 10 minutes prior to counting larvae. 

Fruit Measurements

The fruit temperature in the interior of the berry was measured near weekly in 2019 and 2020 for a random subsample of 0-10 apparently marketable blackberries per replicate plot, depending on fruit availability. Subsamples were averaged to generate one sample data point per plot week prior to analysis, and between 1.4-8.2% of total subsamples were missing due to low yields. Measurements were conducted before harvesting by inserting a thermocouple probe (Digi-sense Handheld Thermometer 86460-06, Cole-Parmer Instrument Co., Vernon Hills, IL) approximately 3 mm into the center of the fruit while it was still on the bush (Schöneberg et al. 2020). Measured fruit were collected separately to avoid categorizing them as damaged/unmarketable fruit and weighed as marketable fruit.

The effect of canopy density on fruit firmness (cN) was measured for subsamples of marketable blackberry fruit. In 2019, depending on fruit availability, 0-5 fruit per replicate were measured after each harvest. In 2020, measurements were conducted only two and three times during the season at WYE and WMREC, respectively. Five drupelets per fruit were measured to account for variation in individual drupes. Measurements were conducted as described in Burrack et al. (2013). Briefly, a flat tipped tension gauge modified with an insect pin was depressed (blunt end) onto the fruit surface at a 90° angle and gentle pressure was applied until the skin was pierced. Subsamples were averaged to generate one sample data point per plot week prior to analysis, and between 10.5-32.2% of total subsamples were missing due to low yields.

Total soluble solids (TSS), measured as °Brix, indicates the amount of sugar and other organic compounds in fruit juice, and higher values correspond to increased sugar content. Between 0-5 marketable berries from each treatment were pooled for a sample, as available. The berries from each sample were crushed together in a WhirlPak^® bag (Nasco, Wisconsin, USA) with a filter to remove particulate matter from the juice. The filtered liquid was transferred to a handheld refractometer (PAL-1, Atago^®, Bellevue, Washington, USA) to determine °Brix and each juice sample was measured in triplicate. Between measurements, the sensor was rinsed with deionized water and dried. Measurements were conducted after each harvest in 2019 and three times during the season in 2020 at both farms. Subsamples were averaged to generate one sample data point per plot week prior to analysis, and between 12.5-28.7% of total samples were missing due to low yields.

Canopy microclimate

HOBOware sensors (HOBO Pro V2, U23-001, Onset Computer Corporation, Bourne, MA) were installed mid-height in the interior (center of the row) of the canopy in the center of each replicate plot to determine treatment impacts on canopy microclimate. The sensors recorded temperature and relative humidity (RH) at 20-minute intervals throughout the fruiting season (July 2019 – September 2019; August 2020 – October 2020). Additional HOBOware sensors (HOBO Pendant^® Temperature/Light, UA-002-64, Onset Computer Corporation, Bourne, MA) were installed to record light intensity (LUX) data at 20-minute intervals throughout the fruiting season on the research farms. Climate parameters were standardized to determine the predictive power of biologically relevant thresholds for D. suzukii survival as described in Schöneberg et al. (2020). Briefly, the number of hours per week (seven days prior to sampling at each site) where temperatures were ≥ 30.9 °C (Ryan et al. 2016) and below 70% RH (Tochen et al. 2016) were calculated. For light intensity, hours <100 LUX (darkness), 100-400 LUX (dusk and dawn), 401-1,000 LUX (overcast daylight), 1,001 – 25,000 LUX (full daylight) and >25,000 LUX (direct sunlight) were calculated. At WMREC, all data loggers were removed during the spray applications.

Spray coverage and deposition

In 2020, spray coverage and spray deposition were evaluated at WMREC. Spray applications were made twice in September (9/11; 9/25) using tartrazine (yellow #5) (1 g L^-1) at a rate of 100 gallons per acre (gpa) using a Durand Wayland 150 gallon airblast sprayer (Model CDP20P150P, Durand-Wayland, Inc., Georgia, USA) with a 81.3 cm fan, a two-way row crop head, and Durand Wayland hollow cone pattern disc spray nozzles. Calibration was performed at the beginning of the season by filling the tank completely, running the sprayer for one-minute intervals, capturing and measuring the spray volume. To achieve 100 gpa, 100 psi of pressure and a speed of 3 mph was used for applications. One side of the sprayer was turned on and driven past each side of each row, with the sprayer passing approximately 0.3m from the exterior canopy foliage. To improve spray coverage, a nonionic surfactant [Latron-B 1956 (1 mL L^-1)] was added.

Spray coverage was evaluated by deploying spray cards (11 x 14 cm white paper, glossy finish, C/2s coating) at three different heights (60 cm, 120 cm and 180 cm above ground) and at alternating locations within the row (2 outside of the canopy and 2 inside the canopy). Spray cards were positioned using 2.5m plastic coated steel core stakes and attached using cable ties. Each stake was placed 0.6m apart from one another. All cards were oriented such that the longest side was horizontal and parallel to the ground. Spray coverage was only quantified on the side facing toward the sprayer (outward toward the row middle). After the tartrazine application was dry (1 h), spray cards were collected and brought to the laboratory. Cards were scanned using an Epson Perfection V 330 at 600 DPI (Epson, Oregon, USA). The outer margin from each card was cropped to standardize the measuring area. The percentage of the card dyed yellow was calculated using ImageJ software as described in Lewis and Hamby (2020). Briefly, thresholding of 0-70 for hue, 17-255 for saturation and the full range 0-255 for brightness was used to create a binary image. Once cards had been converted, ImageJ was used to calculate the proportion of the card stained yellow, providing a quantitative measure of percent coverage.

Spray deposition was measured by collecting fruit one hour after the application. Only green colored unripe fruit were sampled to avoid fruit juice interfering with our measurements. Fruit were sampled from three canopy zones (0-60 cm, 60-120 cm, and 120-180 cm), with up to 5 fruit per zone collected from the exterior of the canopy and the interior of the canopy from both sides of the row (total up to 30 fruit per treatment). Fruit were pooled for sample analysis providing a single data point per replicate plot, location, and height on each application date, and 7.3% of the total 96 samples were missing due to low yields. To avoid touching and inadvertently removing tartrazine from the fruit, scissors were used to collect fruit directly into pre-weighed Whirlpak^® filter bags (Nasco, Wisconsin, USA), which were placed in a chilled container and returned to the lab. Bags were weighed, filled with 5mL RO-water per fruit, and gently shaken by hand for 20 seconds before collecting the resulting solution in 15mL Falcon^® tubes (Corning, Arizona, USA). From there 200µl aliquots of each sample were transferred to a flat bottom 96 well plate. The entire plate was shaken on medium for 5 seconds before quantifying the absorbance at 450nm using a Filter Max F5 Microplate reader (Molecular Devices, San Jose, CA). A standard curve was generated to determine the relationship between concentration and absorbance, as well as the minimum and maximum absorbance values. Serial dilutions were made from a stock solution of 1 g L^-1 tartrazine + 1 mL L^-1 Latron B 1956 generating concentrations of 1 g L^-1, 0.75 g L^-1, 0.5 g L^-1, 0.25 g L^-1, 0.1 g L^-1, 0.075 g L^-1, 0.05 g L^-1, 0.025 g L^-1, 0.01 g L^-1, 0.0075 g L^-1, 0.005 g L^-1, 0.0025 g L^-1, and 0.001 g L^-1. Absorbance of 1 g L^-1 to 0.25 g L^-1 was too high, but absorption values between 0.1 g L^-1 and 0.001 g L^-1 resulted in a linear relationship (R² = 0.999). The standard curve (0.1 g L^-1 – 0.001 g L^-1), water blanks and each sample were measured in triplicates. Unknown concentrations were calculated from known absorbance values and expressed as tartrazine µg (g fruit)^-1.

Data were summarized across subsamples in various ways that are described in the meta data tab of the Excel spreadsheet. Data that were not collected due to low fruit yield or data logger malfuctions are marked not available (NA).