Berry qualities, ripening evenness, and resiliency to mechanical harvest damage

Haskap fruit for this study were sourced from the cold hardy fruit trial at Montana State University – Western Agricultural Research Center (MSU-WARC) in Corvallis, Montana. In 2015, nine plants of each of 15 cultivars (Table 1) were planted in a randomized complete block design, with three blocks and three plants per cultivar per block, totaling 135 plants. Plants were minimally pruned in the late winter months of 2020 to remove broken and low branches to facilitate ease of harvest.

In 2019, one 6-oz (by volume) random sample container of fruit was obtained for laboratory testing following either hand harvest or mechanical harvest “shake-and-catch” from each of two plants per cultivar for genotypes 85-19, ‘Aurora', ‘Blue Goose', ‘Chito', ‘Indigo Gem', ‘Kawai', Solo^TM, ‘Sugar Mountain Blue', 'Tana', and ‘Wild Treasure’ following harvest. Workers were not given special instructions to be gentle in handling fruit. The shake-and-catch harvest method involved shaking individual shrubs with a reciprocating saw equipped with a forked attachment and catching fruit in either a modified plastic kiddie pool or the metal “wings” from the Indigo Waxwing harvester.

Video demonstrating this harvest method:

 

Partway through the harvest season, we observed that “person harvesting” as well as “time of day” were factors contributing to fruit damage, which informed our harvest methods for 2020. Note that a plant growth regulator treatment was applied to one plant per block per cultivar, but these plants were excluded from the 2019 experiments and analyses presented here.

In 2020, haskap fruit assessment was designed as a completely randomized experiment with cultivar as the variable. Sixteen fruit from each of the nine plants for each cultivar in the field trial were hand harvested; workers attempted to collect fruit representative of the whole plant (from all sides, upper and lower, interior and exterior portions of the canopy). Based on our experience from 2019, we modified our hand harvest method to utilize careful and consistent workers and scheduled harvests during the cool morning hours between 6 and 7 a.m. to better preserve fruit quality. Harvested fruit from all plants of each cultivar were pooled in a collection bin for a total of approximately 144 fruit per variety. Fruit were then selected at random from the bin for subsequent lab measurements.

For all cultivars in 2019 and 2020, we collected random samples of harvested fruit and recorded weight (g), length (mm) and width at the thickest part of the fruit perpendicular to the stem (mm); retention of dried flowers and/or leaves (Figure 1); and stem attachment characteristics (wet, dry, or stem attached). Skin elasticity (N) and Brix or flesh firmness (N) and Brix for individual fruits were measured on separate sets of fruit since these procedures destroy fruit. Juice pH and titratable acidity were also recorded. More details on these procedures are provided below.

In 2019 trials, we documented haskap dry matter content for all cultivars by weighing fruit in sets of 10 and drying them in a dehydrator at 150 ˚F until weights stopped fluctuating (n = 6 [sets of ten fruit]). On harvest days, we recorded damage to fruit according to harvest method (Figure 4A). However, we observed that damage became more apparent during storage (Figure 4B); therefore, we added day-after-harvest assessments to 2020 measures of fruit durability.

In 2020, besides the assessments mentioned above, we added a fruit durability test (“drop-and-roll”) in an effort to simulate the dropping and rolling actions that impact berries during mechanical harvest. This procedure involved dropping individual fruit from a height of 7 ft (2.13 m) followed by rolling each fruit down a 10 ft-long (3.05 m) x 4.75 in (12.065 cm) diameter plastic pipe angled at approximately 45°. Each fruit was immediately assessed for damage following the drop-and-roll (“no damage”=0, “damage”=1), placed on trays in individual cells, stored overnight in a cold room cold room (34 ˚F [1 ˚C] / 98% relative humidity [RH] as measured by a Hobo data logger (Onset Computer Corporation)), and reevaluated for damage. As we wanted to control for fruit damaged by hand harvesting versus through the drop-and-roll experiment, a set of fruit that were not dropped-and-rolled was also assessed for damage the day of harvest and similarly placed in cold storage overnight to be reevaluated the following day.

Video demonstrating the berry "drop-and-roll":

Dr. Rachel Leisso first drops a haskap berry from a height of 7 ft., then rolls it down a 10-ft. pipe angled at approximately 45°, and examines the berry for damage.

Berry skin elasticity and firmness were collected using a FR-5120 Digital Fruit Firmness Tester (QA Supplies, Norfolk, VA) equipped with a 2 mm probe mounted on a stand. Measurements were collected in “fast” and “Peak H” mode. This type of device—even when mounted on a stand—can produce variable results due to user technique. Particularly in 2020, we gave more attention to standardizing methods and attempted to utilize the same worker to measure skin elasticity and flesh firmness throughout the trial. Skin elasticity was tested by placing fruit on its flattest side and depressing the penetrometer until the skin was punctured. Fruit firmness was measured on a different set of fruit by placing fruit on its non-flat side (so that one side of the two-berries-within-a-berry would be facing up to the penetrometer as depicted in Figure 2), removing the smallest amount of skin possible to expose berry flesh, and depressing the penetrometer. Brix readings were collected for juice obtained from individual fruit following penetrometer measurements with a Hanna hand-held refractometer (HI 96822, Hanna Instruments).

Sufficient juice volume for measuring titratable acidity and pH was achieved by pooling ten fruit per cultivar for each of the fruit sets tested for skin elasticity or flesh firmness. The pooled fruit were placed in small Ziploc^TM bags and macerated by hand to release the juice. A corner of the bag was removed so that juice could be drained into several 2 mL centrifuge tubes and centrifuged at 13,000 x g to deposit sediment. The supernatant was decanted and used in pH and titration on a Hach AT1000 potentiometric titrator (Model AT1102) equipped with an autosampler (AS1000) and using the pH IntelliCAL^TM PHC281 pH Ultra Refillable pH electrode (Hach Inc., Loveland, CO). Juice samples were titrated to pH 8.2 with 0.25 M NaOH. Results are expressed in g citric acid per L of solution, as citric acid is the predominant acid in for haskaps (Rupasinghe et al., 2018).

Since we neither recorded nor accounted for the confounding factors of “worker harvest technique” and “time of day” factors throughout the 2019 season, all recorded damage was attributed to harvest method (excluding the factor genotype or cultivar). We then tabulated results in contingency tables (number of fruit in the “damage” versus “no damage” condition) and analyzed for “goodness of fit” according to chi-square distribution, where the null hypothesis is that there is no difference between the observed (damage versus no damage) and expected (damage versus no damage) frequencies. Fruit dry matter content is reported as percent dry matter according to cultivar.

For 2020, fruit damage data were analyzed with a binary logit model (Fisher’s scoring optimization technique) via PROC LOGISTIC in SAS. Where Wald’s chi-square test statistic indicated cultivar significance; follow-up cultivar comparisons were performed with pairwise contrast statements. A comparison of day-of-harvest versus day-after-harvest damage was analyzed with the chi-square test statistic in SAS via PROC FREQ. We also performed partial least squares discriminant analysis (PLS-DA) via MetaboAnalyst (Chong et al., 2019). Direct correlations of the fruit’s physical and qualitative parameters were not possible due to mutually exclusive destructive sampling (e.g. skin elasticity, fruit firmness, Brix, and fruit drop-and-roll). Therefore, we modeled percent fruit damaged for each cultivar in ranges of 0-25%, 26-50%, 51-75%, and 76-100%.

Flower retention, stem scar wetness, and stem and leaf attachment were analyzed via chi-square analysis. Where the Chi-square test statistic was significant at p <0.05, post-hoc tests with Bonferroni’s adjustment for multiple testing comparison were performed. Chi-square test statistics are reported below the table in the form (degrees of freedom, sample size) = chi-square test statistic and p-value.

Length, width, weight, skin elasticity, flesh firmness, Brix, pH, and TA were analyzed using the Statistical Analysis System (SAS) (version 9.4, SAS Institute, Cary, NC). Assumptions of normality were assessed using the Shapiro-Wilk test statistic performed with PROC UNIVARIATE. Where testing several transformations did not correct data to meet the assumption for normality, the non-parametric Kruskal-Wallis test was performed. Homogeneity of variances were tested per Levene’s test; if cultivar variances were non-homogenous, then Welch’s ANOVA (unknown) was used. Tukey’s post-hoc test was used to establish means separation and is equivalent to the Games-Howell test (2019) which does not require homogeneity of variance. Ridge plots were produced in R (Team, 2017) using packages ggplot2 (Wickham, 2009) and ggridges (Wilke, 2021).