Experimental site.

The three watermelon cultigen evaluations were conducted during a field experiment at the Central Crops Research Station (lat. 35.6507°N, long. −78.4962°W) located in Clayton, NC (North Carolina State University) from May to September in 2012, 2013, and 2014. The soil type was sandy clay loam [80% sand, 5% clay, low organic matter (1% to 1.5%), and low acid retention] (Kleiss, 1981). The average maximum and minimum temperatures, rainfall, and cumulative rainfall for May to September in 2012, 2013, and 2014 are provided in Table 1. Day and night temperatures and daily rainfall during peak pollination and fruit set (3–6 weeks after transplanting) are reported in Fig. 2.

Table 1.Average day and night temperatures, minimum and maximum temperatures, and average daily and cumulative rainfall for each cultigen evaluation year from May to September.

View Fullscreen

View Figure

• Download as Powerpoint
• Download Figure

Experimental design and treatments.

A randomized complete block design (RCBD) with four replications was used for each watermelon cultigen evaluation in 2012, 2013, and 2014, with five fruit randomly collected from each plot (20 fruit/entry) used as the experimental unit for evaluations. New seed were obtained each year and consisted of 44, 40, and 30 entries for 2012, 2013, and 2014, respectively. Additionally, the common cultigen data set includes the 13 common cultigens that were included in all 3 years. The common cultigens and fruit characteristics are listed in Supplemental Table 1.

Seedling and field preparation.

Seed of triploid cultigens and diploid pollenizer plants were sown 5 Apr. 2012, 5 Apr. 2013, and 15 Apr., 2014. Seeds were sown into polythylene transplant trays (Hummert International, Earth City, MO) filled with commercial soilless mix (Fine Germinating Mix; Carolina Greenhouse and Soil Company, Kinston, NC) and held under greenhouse conditions (range, 24 to 30 °C). Approximately 3 weeks after seeding, watermelon transplants were placed in either a coldframe or a high tunnel to harden. Transplants were field-planted 8 May 2012, 14 May 2013, and 19 May 2014. Granular fertilizer was the same for each year and applied at 33.6 kg·ha⁻¹ N and 74.1 kg·ha⁻¹ K. Granular fertilizers were incorporated into the beds before laying black polyethylene plastic (0.70-mm-thick high-density plastic film, 1.2 m wide; B.B. Hobbs, Clinton, NC).

The study fields were fumigated with Telone C-17 (Dow AgroSciences, Indianapolis, IN) via broadcast application at 93.4 kg·ha⁻¹ on 14 Nov. 2011, and the same rate was applied on 9 Nov. 2012 and 7 Nov. 2013. Ethalfluralin (4.7 kg·ha⁻¹), N,N′-dimethyl-4,4′-bipyridinium dichloride (3.5 kg·ha⁻¹) and bensulide (5.8 kg·ha⁻¹) (Dow AgroSciences) were applied between plastic beds for weed control in May (before planting) and June (before watermelon started to vine out) for all triploid watermelon evaluations (Schultheis and Thompson, 2012, 2013, 2014).

Crop production.

Spacing between rows was 3.1 m, with in-row plant spacing of 0.8 m. Plots were one row and 7.6 m long, allowing 10 plants to be grown for each experimental unit. A 3.1-m alley was maintained between plots. At the time of transplantation, 20–20–20 liquid fertilizer (20.0N–8.7P–16.5K, Peters Professional General Purpose Fertilizer; ICL Specialty Fertilizers, Summerville, SC) and Coragen (DuPont, Wilmington, DE) were applied at 0.01 kg·ha⁻¹ and 0.07 kg·ha⁻¹, respectively. Plots with missing transplants were replanted ≈7 d after planting to achieve a 100% stand count for all three experiments (Schultheis and Thompson, 2012, 2013, 2014).

Pollenizer plants.

During the 2012 growing season, the diploid pollenizer ‘Ace’ was interplanted after triploid plants 1, 4, and 7 in each plot. In 2013, two pollenizer cultigens (‘Ace’ and ‘SP-6’) were used, with ‘SP-6’ interplanted after triploid plants 1 and 10, and ‘Ace’ was interplanted after triploid plants 4 and 7. ‘SP-6’ was used as the sole pollenizer source in 2014; it was interplanted after plants 1, 4, 8, and 10.

Trickle irrigation (NETAFIM, 197 mL, 1.09 × 10³ mL·h⁻¹; NETAFIM, Tel Aviv, Israel) was used over the course of each growing season. Fertigation was started 2 weeks after planting and was drip-applied weekly over the course of the growing seasons using a 4–0–8 (4.0N–0.0P–6.6K) liquid fertilizer (Harrell’s Max Potassium plus Calcium; Harrell’s Inc., Lakeland, FL). Cumulative amounts of liquid fertilizer applied throughout the growing seasons were 102.2 kg·ha⁻¹ N, 0.0 kg·ha⁻¹ P, and 166.9 kg·ha⁻¹ K.

A conventional spray program for North Carolina was used for watermelon crop production (Schultheis and Thompson, 2012, 2013, 2014). Insecticide, fungicide, and/or miticide applications were applied every week starting from the first week of June through the second week of August. The products were alternated during consecutive spray applications among the evaluations to avoid pest resistance. Acramite (Arysta, Tokyo, Japan), a miticide, was applied in 2012; the fungicide and insecticide programs used during each of the growing seasons are listed in Supplemental Table 2.

Crop harvest and data collection.

Four watermelon harvests were performed in 2012 and 2014, and five harvests were performed in 2013. The first harvests were performed on 17 July 2012, 18 July 2013, and 24 July 2014. The fruit were harvested at ≈65 d after anthesis when ripe, weighed, and rated for the incidence and severity of HH. Using a scale of 0 to 4, HH was rated in 2012; 0 indicated no HH, 1 indicated slight HH, and 4 indicated severe HH. During the 2013 and 2014 evaluations, a scale of 1 to 5 was used; 1 indicated no HH, 2 indicated slight HH, and 5 indicated severe HH (USDA, 2006). For uniform comparisons across years, HH ratings for 2012 were converted to a scale of 1 to 5. Other evaluations for each entry included yield, fruit shape and size, rind pattern, occurrence of hard seeds, flesh color, and soluble solids content (NC State Extension website, https://cucurbits.ces.ncsu.edu/growing-cucurbits/variety-trials/).

Flesh firmness in pounds was determined using a Penetrometer FT 011 (range, 1–11 lb/4.4–44 N) with a 1.11-cm-diameter plunger tip (QA Supplies LLC, Norfolk, VA). Firmness readings were taken at five locations in the placental material (flesh) (Fig. 3) at the ground spot side, opposite (top) side, blossom end, stem end, and heart; they are presented as means for 20 fruit per cultigen per year, primarily from the first or second harvests. Fruit firmness readings were converted from pounds to Newtons.

View Figure

• Download as Powerpoint
• Download Figure

Data analysis.

All statistical analyses of HH percent (incidence), severity, and tissue firmness were conducted using SAS 9.4 (SAS Institute, Cary, NC). The large multiyear dataset was analyzed by year because each year had a large number of cultigens not included in the other years. Data for the 13 common cultigens were analyzed by year and across the years. Linear regression was used to determine if there was a linear relationship between tissue firmness and incidence of HH. Coefficient of determination (R²) values were also used to assess the relationship between tissue firmness and incidence of HH. An analysis of variance (ANOVA) was used to quantify cultigen differences in tissue firmness and HH severity. Logistic regression was used to evaluate the incidence of HH (HH%) and calculate cultigen predictive odds ratios for exhibiting HH when compared with ‘Maxima’. Mean separations were performed using the Tukey-Kramer honestly significant difference test (P ≤ 0.05) for average tissue firmness (N), average HH severity rating, and HH incidence among the large multiyear cultigen evaluations and common cultigens.

Incidence and severity of HH.

Logistic regression indicated that the incidence of HH did not differ by year during the large multiyear cultigen evaluation (Supplemental Table 3) (P > 0.05). Within years, the range of HH incidence was highest in 2012 (0% to 65%), lowest in 2013 (0% to 35%), and intermediate in 2014 (0% to 50%) (Supplemental Fig. 1). During the large multiyear cultigen evaluation, the severity of HH did not differ by year or across the years (data not shown) (P > 0.05).

For the 13 common cultigens, the interaction of year × cultigen was not significant (P > 0.9567). The incidence of HH did not differ for the 13 common cultigens within years (P < 0.1221, P < 0.3387, and P < 0.2327 in 2012, 2013, and 2014, respectively). The 3 years of data for the 13 common cultigens were combined and subjected to logistic regression, which resulted in significant cultigen differences in the incidence of HH (P < 0.0001) (Table 2). The 13 common cultigens did not differ in HH severity within or across years (data not shown) (P > 0.05).

Table 2.Incidence of hollow heart (HH) and predictive odds of a cultigen in the common cultigen data set for developing hollow heart in any given year.

View Fullscreen

Tissue firmness and hollow heart.

The mean tissue firmness found during the large multiyear cultigen evaluation differed by year (P < 0.05) (Supplemental Fig. 1). In 2012, tissue firmness ranged from 11.0 to 17.6 N, with the lowest values for ‘Gilboa’ and ‘Affirmed’ (11.0 N and 11.2 N, respectively) and the highest value for ‘Fusion’ (17.6 N). In 2013, tissue firmness ranged from 10.3 to 16.2 N, with the lowest value for ‘Tri-X-313’ (10.3 N) and highest value for ‘LaJuva’ (17.2 N). Tissue firmness ranged from 9.1 to 16.4 N in 2014; it was lowest for ‘Harvest Moon’ (9.1 N) and highest for ‘Premont’ (16.4 N), ‘Captivation’ (16.3 N), 13-13077 (16.1 N), ‘Crunchy Red’ (16.0 N), ‘Exclamation’ (15.6 N), ‘Secretariat’ (15.3 N), and ‘ACX 6177’ (15.3 N).

Coefficients of determination indicated that the incidence of HH was positively correlated to the severity of HH (0.96; R² = 0.92), and the mean fruit tissue firmness was negatively correlated to the incidence of HH (0.55; R² = −0.32) (data not shown). In 2012 and 2013, cultigens that had the lowest mean tissue firmness (‘Affirmed’, ‘Gilboa’, and ‘Tri-X-313’) were in the top 25% of cultigens for HH incidence. Conversely, cultigens with the highest tissue firmness (‘Fusion’ and ‘LaJuya’) were in the lowest 10% of cultigens for incidence of HH. No trends were found in 2014 for cultigen flesh firmness and HH incidence.

For the common cultigen set, mean tissue firmness differed across the years (P > 0.0001) (Fig. 4), and cultigens with higher tissue firmness values had a lower HH incidence. Cultigens with an average HH incidence of 20% or less, such as ‘Crunchy Red’, ‘Maxima’, and ‘Captivation’, had firmness values of 15 to 16 N. Conversely, tissue firmness was 11 to 12 N for four of six cultigens with an HH incidence more than 20%. The mean tissue firmness was negatively correlated to the incidence of HH (0.88; R² = −0.78). The mean tissue firmness was also negatively correlated to HH severity (0.62; R² = −0.39). In contrast, HH incidence and severity was positively correlated (0.96; R² = 0.92).

View Figure

• Download as Powerpoint
• Download Figure

Percentage of unmarketable fruit.

When the 13 common cultigens were combined across years, the percentage of unmarketable fruit (HH > 2) differed among cultigens (Fig. 5). However, the percentage of unmarketable fruit for the 13 common cultigens did not differ within 1 year or interact with year (P > 0.05). ‘Liberty’, ‘Bold Ruler’, and ‘Declaration’ had the highest percentage of unmarketable fruit (28.3%, 23.3%, and 21.6%, respectively), whereas ‘Captivation’ had the lowest percentage of unmarketable fruit (5%).

View Figure

• Download as Powerpoint
• Download Figure

Linear regression.

In the common cultigen set, linear regression was used to determine the relationship between tissue firmness and HH incidence (%) (Fig. 6). Fruit with lower tissue firmness generally had a higher percentage of HH. ‘Bold Ruler’, ‘Liberty’, and ‘Affirmed’ had low tissue firmness (≈12 N) and a high probability of developing HH. Conversely, ‘Maxima’ and ‘Captivation’ had the lowest probability of developing HH and higher tissue firmness (15 N). Using average tissue firmness for 2012–14 in the common cultigen data set, the incidence of HH was predicted using the following formula: y = −6.6152(N) + 115.29, where y = %HH and N represents firmness (Newton). The general trends of higher firmness and less HH incidence also indicate that selection of cultigens for higher firmness may reduce HH incidence and severity.

View Figure

• Download as Powerpoint
• Download Figure

Because HH incidence is unpredictable within a growing season, other factors that might contribute to this disorder have been explored. Although HH disorder is found in diploid (seeded) fruit, it is more prevalent in triploid watermelons, which do not produce sufficient viable pollen for fruit set. Peak flowering and pollination of watermelon generally begin at ≈2 to 3 weeks after field transplanting and can continue for an additional 6 weeks (Dittmar et al., 2009; Freeman et al., 2007). Watermelon pistillate flower stigmas are most receptive to pollen from 15 to 26 °C and senesce within a few hours (Lyu et al., 2019). Pollinators need to visit a triploid flower 16 to 24 times to achieve adequate pollen grain transfer and fruit set (Fiacchino and Walters, 2003; Walters, 2005). Pollen amount and viability, pollinator activity, and pistillate flower receptivity are highly influenced by environmental factors such as rainfall and fluctuations in day and night temperatures, and they contribute to HH formation (Fiacchino and Walters, 2003; Freeman and Olson, 2007a; McGregor and Waters, 2014). High daytime temperatures (>30 °C) and low morning temperatures (>15 °C) or fluctuations in humidity soon after anthesis can decrease pollen viability (Lyu et al., 2019) and trigger the onset of HH (Stanghellini and Schultheis, 2005).

The use of increased ratios of diploid to triploid plants in this study (43%) instead of commercial recommendations (25%) (Fiacchino and Walters, 2003; Freeman and Olson, 2007b) should have reduced potential pollen deficiencies that may cause HH. In addition, using pollenizers with varied timing of flowering should provide ample pollen throughout production and help reduce HH incidence. ‘Ace’ has been reported to have an early flowering date and elongated pollination period, making pollen available for the first female triploid watermelon pistillate flowers as well as those produced later (Sakata Vegetables, 2018). In comparison, ‘SP-6’ has high pollen release/quantity, thin branches, and small, deep lobed leaves to increase flower visibility to honeybees over an extended period of time (Brusca and Zhang, 2012). Although first bloom dates for male or female flowers were not recorded during the watermelon cultigen evaluations used for this study, male flowering usually begins at 10 to 14 d posttransplant, and female flowers appear at 21 d posttransplant (McGregor and Waters, 2014). Average day and night temperatures were similar during the three growing seasons (May–September), although cumulative rainfall was considerably higher in 2013 compared with 2012 and 2014 (Table 1). The large rainfall event in 2013 occurred near the end of the watermelon season, making it less likely to contribute to HH.

Despite the increased ratio of diploid pollenizers, HH still occurred in this study, indicating that some germplasms appear to be more susceptible to HH. In the common cultigen set, ‘Liberty’, ‘Bold Ruler’, and ‘Affirmed’ consistently showed a higher HH incidence (Fig. 5). These cultigens may differ from other cultigens in earliness or lateness of bloom (before adequate pollen is available), or flowers may require more pollen than other cultigens to set fruit and grow normally.

Kano (1993) postulated that when the watermelon placental tissue growth rate (e.g., rapid and irregular fruit growth) exceeds a certain point, tension is imposed on the cells, causing middle lamella separation and HH disorder. The common cultigen set indicates that a higher HH incidence is associated with lower tissue firmness (Figs. 4 and 6). Linear regression of the incidence of HH and mean fruit firmness in the large multiyear cultigen evaluations (data not shown) and the 13 common cultigen set provided negative regression between fruit firmness and HH incidence (Fig. 6). This suggests that fruit with lower tissue firmness are more susceptible to HH.

Logistic regression.

When a dependent variable is considered binomial, logistic regression provides a more robust statistical comparison than linear regression or ANOVA (Lever et al., 2016). Logistic regression was used to determine the incidence of HH (HH%) in the large multiyear data set and common cultigens. Logistic regression was performed separately by year for the large multiyear cultigen set because the cultigens varied considerably across years. No differences were found among the multiyear cultigen evaluations for HH incidence (Supplemental Table 3), probably because of the high amount of variation within and across replicates. A difference was observed among cultigens in 2012 (P < 0.003), but there was no difference in 2013 (P < 0.300) or 2014 (P < 0.200). Cultigen differences in 2012 likely occurred because there was a higher incidence of HH in 2012 than in the other years.

The 13 common cultigen data set was analyzed by year and across years with data combined. When combining data across years, the common cultigens showed differences in HH incidence (Table 2). Logistic regression for the common cultigens also provided the predictive odds of developing HH when compared with ‘Maxima’, which was one of the cultigens that was least likely to develop HH. The predicative odds indicated that ‘Bold Ruler’ and ‘Liberty’ were 7-times and 6.9-times more likely, respectively, to develop HH than ‘Maxima’, which had one of the lowest odds of developing this disorder. ‘Captivation’ was the only cultivar with lower odds than ‘Maxima’ and was 0.4-times as likely to develop HH as ‘Maxima’ (Table 2).