Breeding Methods

by Todd C. Wehner
Department of Horticultural Science
North Carolina State University
Raleigh, NC 27695-7609

Breeding Objectives

Major objectives for watermelon breeding include proper fruit type, early maturity, high fruit yield, high sugar content, tough flexible rind, and proper seed type. It is important to determine breeding objectives carefully before starting variety development. For example, seed type changes significantly for different market classes. Parental lines for seedless hybrids should have small seeds, whereas confectionery seed types should have large seeds. For commercial varieties, black seeds are preferred because of their contrast with red, yellow, or orange flesh. Also, white seeds indicate immaturity to buyers, so white mature seed color can be a confusing trait for them. Most of the old varieties are diploid, open-pollinated or inbred lines, but hybrid diploid and hybrid triploid varieties are taking over the commercial market in the United States.

After determining the breeding objectives, methods for measurement of the traits of interest should be developed, selection methods should be determined (specifying the operations to be carried out for each generation), and parents with high expression of the traits of interest should be chosen. Vine type should be long for commercial production and dwarf (bush) for home garden. It may also be possible to use the dwarf plant type for once-over harvest in commercial production. Sex expression should be monoecious, with a ratio of 7 staminate:1 pistillate flowers, or better (preferably 4:1). Andromonoecious sex expression and ratios of 15:1 are more typical of older varieties.

For production in most areas of the United States, watermelon must have resistance to fusarium wilt. Races 0 and 1 are common, and race 2 is becoming important, especially in Texas and Oklahoma where plastic mulch culture and fumigation are less common. Production areas in the southern United States usually have anthracnose race 1 and may also have problems with race 2. Gummy stem blight is a disease for which resistance is needed in most southern production areas. Powdery mildew is becoming a problem, especially in the western United States (possibly because of a new race), and should be a breeding objective for new varieties. Bacterial fruit blotch was a problem in the 1990s, and resistant accessions have been identified. The disease can be effectively controlled by genetic resistance and by large-scale seed testing followed by destruction of contaminated seed lots. Protection from viruses in the United States production areas should include resistance to papaya ringspot virus-watermelon strain (formerly watermelon mosaic virus-1), watermelon mosaic virus (formerly watermelon mosaic virus-2), and zucchini yellow mosaic virus.

Finally, breeding objectives should emphasize early maturity, high fruit yield, durability for shipping, high internal quality, freedom from internal defects (hollowheart and rind necrosis), and proper seed type in a diploid (seeded) or triploid (seedless) hybrid. Internal quality traits include dark red flesh, high sugar content, proper sugar to acid ratio, excellent flavor, high nutritional value (vitamins and lycopene), firm (not soft) and non-fibrous texture. Seeds should be black color, medium size (or small for inbreds to be made into tetraploids), and few to medium quantity per fruit (few for consumers, but medium to keep seed costs down). Flesh color should be dark red (Y gene with modifier genes) with uniform color throughout the fruit. For specialty types, flesh color could be bright orange (yO gene), canary yellow (C gene), or white (Wf gene). Other colors such as salmon yellow (y gene) exist, but are not preferred because the flesh looks overmature. Older varieties have light red flesh, but dark red is becoming the preferred type. Diploid inbreds should be made into tetraploid inbreds and tested for fertility, seed yield, and ability to set fruit using controlled pollination. Tetraploid lines for use in triploid seedless hybrid production can be induced with colchicine. Finally, triploid hybrids should be tested for absence of seed coats in the fruit within a range of production environments.

Variety Development

There were no defined varieties of watermelon before the 1820s. Early varieties include ‘Black Spanish’ (imported to United States from Portugal in 1827), ‘Carolina’ (available at least since 1827), and ‘Imperial’, ‘Mountain Sprout’, ‘Seminole’, and ‘Mountain Sweet’ (introduced by southern growers from 1840 to 1850). Other heirloom varieties include ‘Bradford’, ‘Clarendon’, ‘Odell’, ‘Ravenscroft’, and ‘Souter’ (originating in South Carolina before 1850). Classic watermelon varieties include ‘Peerless’ or ‘Ice Cream’ (1860), ‘Phinney Early’ (1870), and ‘Georgia Rattlesnake’ developed by M.W. Johnson in Atlanta, Georgia about 1870.

Planned variety development programs began in the United States in 1880 to 1900. Important varieties developed for the southern United States included ‘Cuban Queen’ developed and marketed by Burpee in 1881, ‘Round Light Icing’ (1885), ‘Kolb Gem’ developed by Reuben Kolb of Alabama in 1885 and marketed by D.M. Ferry, ‘Florida Favorite’ selected from the cross of ‘Pierson’ x ‘Georgia Rattlesnake’ by Girardeau in Monticello, Florida in 1887, ‘Dark Icing’ developed in 1888 by D.M. Ferry, and ‘Dixie’ selected from the cross of ‘Kolb Gem’ x ‘Cuban Queen’ or ‘Mountain Sweet’ by George Collins in North Carolina and marketed by Johnson and Stokes. Important varieties developed for the western United States included ‘Chilean’ (black or white seeded) brought from the west coast of South America and introduced to California in 1900, ‘Angeleno’ developed by Johnson and Musser in Los Angeles, California in 1908, and ‘Klondike Solid’ and ‘Klondike Striped’ of unknown origin developed about 1900. Important varieties developed for shipping include ‘Tom Watson’ developed by Alexander Seed Co. in Augusta, Georgia in 1906, and ‘Stone Mountain’ developed by Hastings Co. in Atlanta, Georgia in 1924.

Important varieties developed in the latter part of last century have built on past accomplishments. ‘Charleston Gray’ (USDA, Charleston, 1954), ‘Crimson Sweet’ (Kansas State University, 1963), ‘Calhoun Gray’ (Louisiana State University, 1965), and ‘Dixielee’ (1979), ‘Jubilee’ (1963), and ‘Smokylee’ (1971) (all from the University of Florida) have high resistance to Fusarium wilt. ‘Dixlee’ (University of Florida, 1979) and ‘Sangria’ F1 (Novartis, 1985) have dark red flesh. ‘Millionaire’ F1, 3x (Harris Moran, 1992) and ‘Royal Jubilee’ F1 (Seminis) have consistently high yields. ‘Crimson Sweet’ (Kansas State University, 1963) and ‘Sugarlee’ (University of Florida, 1981) have high soluble solids. ‘Kengarden’ (University of Kentucky, 1975) has dwarf vines. ‘Tri-X-313’ F1 3x (Novartis, 1962) is seedless. ‘Minilee’ (University of Florida, 1986), ‘Mickylee’ (University of Florida, 1986), ‘New Hampshire Midget’ (University of New Hampshire, 1951), ‘Sugar Baby’ (M. Hardin, Oklahoma, 1955), and ‘Tiger Baby’ (Seminis) are icebox size. ‘Yellow Doll’ (Seminis, 1977) has canary yellow flesh.

A recent trend has been to develop seedless cultivars that have mini fruit size, or 3.2 kg (7 lb) per fruit. It may be of interest to go even small to micro fruit size, or 1 to 2 kg (0.4 to 0.9 lb) per fruit. A micro greenhouse watermelon could even be developed, similar to the Dutch greenhouse cucumber that has been popular for decades. Traits are already available to develop such a cultivar, which could be grown in a greenhouse, with plants supported on a trellis. For example, the following traits (cultivar source in parentheses) might be combined to achieve that objective:

  • Micro size fruit, 0.7 kg each (New Hampshire Midget)
  • Gynoecious sex expression (mutant from China)
  • Optional: non-lobed leaves (Sunshade)
  • Optional: dark solid skin color (Mountain Hoosier)
  • Optional: scarlet red flesh (Dixielee)
  • Elongate fruit shape (Florida Favorite); 300 mm x 100 mm fruit dimensions
  • Thin, explosive rind (New Hampshire Midget)
  • Triploid hybrid (4X-2X F1)

Breeding Plan. Once the breeder has determined the objectives of the program, the choice of parental materials is one of the most important aspects of a breeding program. Using knowledge of the crop and predicting the traits consumers will be interested in having in future varieties, the breeder gathers parental lines for crossing. The breeder should know which parent will contribute the traits of interest, and which methods will be used to evaluate the progeny for those traits. Thus, it is often necessary to collect and evaluate large numbers of PI accessions, varieties, and breeding lines for the traits of interest to identify appropriate parents to use in the program. This work often continues in parallel with the main part of the breeding program.

The next step is to determine the breeding method to use for each part of the program. It is important for the breeder to consider the advantages and disadvantages of particular breeding methods, and how they can be incorporated into the overall breeding plan. Also, it is common to use more than one breeding method at a time in order to accomplish several sets of objectives. For example, one part of the program might be to use recurrent selection to develop a base population with general adaptation and the proper fruit type that also has high yield and early maturity. A second part of the program might be to use pedigree selection on the cross of two lines to develop inbred lines with the high yield, early maturity, and proper fruit type of one parent, and the dark red flesh color, high sugar content, and firm crisp flesh texture of the other parent. A third part of the program might be to use backcross breeding to make a canary yellow flesh version of an elite red-fleshed hybrid with top performance.

Recurrent Selection. Although watermelon is a cross-pollinated crop, population improvement methods popular in some cross-pollinated crops have not been used. The main reason for that appears to be the large size of the plants, and the low rate of natural outcrossing that occurs. Also, because there are few plant breeders working on watermelon, and because of the requirement for many qualitative traits to be present in the new varieties being tested for release, it is expensive to spend additional years in population improvement for quantitative traits.
It may be possible to improve quantitative traits such as yield in watermelon using recurrent selection i.e. repeated selection and massing of selected plants, but the populations should probably be developed initially to have the necessary qualitative genes in them. Those would include proper flesh color, fruit size, and disease resistance. Due to large plant size and a 5-month generation time, recurrent selection methods should be those that have few generations per cycle, and few plants per family (or single-plant selection).

One approach would be to develop an elite population by intercrossing two to four of the best red fleshed hybrids available, trying to choose a set that was genetically unrelated. A population with a wide genetic base could also be developed by intercrossing 20 or more elite varieties by hand for two or more generations, and using bees in an isolation block for two or more generations before beginning a mild selection pressure for important quantitative traits such as yield. Simple recurrent selection could be used for selection among single-plant hills for a set of highly heritable traits. A more complex method such as reciprocal recurrent selection would permit simultaneous improvement of two populations for combining ability for yield. This would be an expensive program to run, but would produce two populations that could be used to develop inbreds to be used as the female and male parents (respectively) of elite hybrids.

During population development, it would be necessary to identify methods for yield testing that were efficient for use in large yield trials. The usual guidelines for recurrent selection are to test at least 200 individuals (or progenies of individuals) per population, and to select at least 20 to intercross for the next cycle of selection. A yield trial involving 200 replicated families would require more resources than many breeding programs could afford if the trial were done using current methods.

Recurrent selection could be used to improve quantitative traits, such as yield, which are difficult to improve using qualitative methods such as pedigree and backcross breeding. Each year, the improved population would be used to begin the development of inbred lines to feed into other parts of the breeding program.

Pedigree Breeding. Probably the most common method for watermelon breeding is pedigree. In pedigree breeding, the breeder begins by choosing two or more adapted parents, which complement each other in their traits. For example, one parent might be generally good (yield, earliness, type) except for disease resistance and the other might be generally good (yield, earliness, type) except for fruit quality. The objective would be to produce new lines with high yield, early maturity, proper type, high fruit quality, and good disease resistance. The varieties or breeding lines are crossed to form the hybrid (F1) generation, which is then self- or sib-pollinated to form a segregating (F2) population (Fig. 3.10). The F2 is self- or sib-pollinated while selecting for traits having high heritability to form the F3 generation. If multiple plants are tested from each selected F2 plant, then the breeder concentrates on selecting the best plants in each of the best F3 families. This might include selection in the seedling stage in the greenhouse in the F2 and F3 generations for disease resistance such as fusarium wilt races 0, 1, and 2 and anthracnose races 1 and 2.

Beginning at the F4 generation, selection would begin to emphasize family-row performance for quantitative traits. Plants within family-rows that have excellent performance for qualitative traits should be selected for the next generation. As the families reach six generations of self-pollination (S6 or F5), they become more uniform, and can then be handled as inbred lines. This could include selection using eight-plant plots for early flowering, number of pistillate flowers, and fruit number. The number handled might decrease from 54 F2 plants of a cross to 36 F3 families, 24 F4 families, and 18 F5 lines.

Single-seed-descent is a modification of pedigree breeding in which inbred lines are developed rapidly by self-pollination in greenhouses and winter nurseries, and selection is not practiced until later generations, such as S3 to S6. This method requires less record keeping and works better where the main objective is to improve quantitative traits such as yield and earliness, rather than qualitative traits such as flesh color and disease resistance. However, traditional pedigree breeding is probably the more useful method for watermelon since there are many qualitative traits that can be selected in early generations. In that way, plants or families having unsuitable traits that are simply inherited (such as poor fruit flesh color) can be eliminated in early generations. Otherwise, they would be carried along until the S3 to S6 generation when field-testing would be practiced in the single-seed-descent breeding method.

Backcross Breeding. Backcross breeding is used to transfer one qualitative (highly-heritable) trait into an otherwise superior inbred. The superior inbred is referred to as the recurrent parent. Often, six generations of selection and backcrossing to the recurrent parent are used to recover the genotype of the recurrent parent (except for the addition of the new trait) without the other undesirable traits from the non-recurrent (donor) parent. Two versions of the backcross method are used depending on whether the gene of interest is recessive or dominant.

For the transfer of a trait controlled by a recessive gene, the recurrent parent is crossed with the donor parent, and the F1 backcrossed to the recurrent parent. In one scheme, the F1 is self-pollinated to produce the F2, which will segregate for the trait of interest. Individuals having the trait can then be backcrossed to the recurrent parent to produce the BC1. The BC1 generation is then tested for the trait, and individuals having it are self-pollinated once again to produce a segregating generation for selection and backcrossing to the recurrent parent. The process is repeated until the BC6 generation when the best individuals are self-pollinated and selected for the trait to produce the improved inbred. The inbred does not need to be tested extensively in trials, because it will be identical to the original inbred, but with one new trait.

For the transfer of a trait controlled by a dominant gene, the recurrent parent is crossed with the donor parent, and the F1 backcrossed to the recurrent parent. The BC1 generation is then tested for the trait, and individuals having it are backcrossed to the recurrent parent. The process is repeated until the BC6 generation when the best individuals are self-pollinated and selected for homozygous expression of the trait using progeny testing.

Inbred Development. The best selections from the recurrent selection program should be self-pollinated each cycle to begin inbred development. Pedigree selection, and backcross breeding result in the production of elite inbred lines. Each year, those inbred lines that are produced from the different parts of the breeding program should be increased by self-pollination, tested for useful horticultural traits, and used in the production of tetraploid inbred lines, as well as directly for the production of diploid hybrids based on the traits they have, and what is needed by the market.
Isolation blocks or screen cages can be used to make large seed increases of the inbreds if that is needed. Isolation blocks should be away from other watermelon fields, requiring a separation of at least 1 mile. Bees should be provided in the isolation block or cage by bringing in one strong hive, unless there are sufficient numbers of wild bees.

Hybrid Testing. The final stage of breeding is to produce hybrids for testing. Hybrids are usually made between two monoecious inbreds. For triploid hybrid production, the seed parent should have a distinctive rind pattern that has recessive inheritance. For hybrid production with less labor input, the seed parent could be male sterile. The seed increase of the male sterile inbred would be accomplished by pollinating male sterile plants with the heterozygote (Ms ms) as the pollen parent. For seedless hybrid production, the seed parent would be a tetraploid inbred.

Once they have been developed, all inbreds can be crossed in all possible combinations. However, that might produce too many entries to evaluate properly. For example, 20 inbreds could produce (20 x 19)/2 = 190 different hybrids, without including reciprocals. Thus, it may make more sense to make hybrids only from pairs of inbreds having complementing traits of the proper type.

Testing of experimental hybrids should progress in stages, with fewer hybrids to test in later stages where more effort is spent on each hybrid. The first year trials might have two replications in each of two locations. In the second year, the best hybrids could be evaluated in 8 to 12 locations using the conditions available at each (grower fields, state university experiment stations). In the third year, the hybrids would be sent to grower trials throughout the production regions of interest for trials involving 0.25 to 1.0 acre using a total of 5-10 lb. of seeds for all trials. Seeds should be screened for bacterial fruit blotch before sending to growers. One can usually get good data from at least 10 of the 50 trials. Information from the 3 years of trialing should lead to the release of the best one or two hybrids in the fourth year.

Although there is not much advantage of hybrids over open-pollinated varieties for most traits, it is thought that the former are more uniform. Thus, it may be possible to get the same yield in fewer harvests because of more uniform growth and a more concentrated fruit set. Hybrids offer several advantages over open-pollinated varieties. A major advantage is the production of seedless triploids, which are produced by crossing a tetraploid female inbred with a diploid male inbred. Hybrids also can express heterosis, with the hybrid performing slightly better than the best parent in some cases. The amount of heterosis in watermelon is around 10%. Another advantage is the ability to get an intermediate fruit shape by crossing an elongate-fruited inbred with a round-fruited one. Inbreds can be used to combine dominant genes for resistance from each parent into a hybrid that has more dominant genes expressed than either parent. A hybrid that has large seeds for the grower to plant and small seeds in the fruit sold to the consumer can be produced by crossing a large-seeded female inbred with a small-seeded male inbred. Finally, hybrids provide a way for the seed company to protect their proprietary inbreds from theft.

The disadvantages of hybrids are that they add an extra step to the breeding process, and increase the cost of seeds since they are produced by hand pollination rather than by bee pollination. Use of male sterile inbreds for seed production should help reduce the cost of hybrid seeds in the future.


The job of watermelon breeding can be made easier and more efficient if mechanization is used for as many steps in the process as possible. Small plot equipment can be used for fieldwork to permit more germplasm to be tested with fewer workers and at a lower cost. Small-plot seeders can be used to plant seeds in the field with optimum seed spacing and planting depth using fewer workers than if seeds are planted by hand. If transplants are used to plant the test plots, machine transplanters can be used to punch the hole before the workers on the machine set the seedling into the hole, and follow up with water and fertilizer after the worker has pressed soil around the seedling, all while riding down the field row. Seeds can be packeted using a seed counter, and plot size can be optimized to gain the maximum information for the lowest cost. Research indicates that optimum plot shape is rectangular and block (replication) shape is square. It is difficult to mechanize harvest since it is done by hand, and each fruit is counted and weighed. However, some efficiency can be gained by using portable computers to collect and analyze data. In the advanced trials, it is useful to estimate flesh sweetness (fruit soluble solids content) using a refractometer, and rind toughness using a spring-loaded punch or penetrometer.

If a greenhouse generation is used to expedite inbred development or hybridization, automation systems are useful for handling the many plants to be grown for self- or cross-pollination. Such systems include automatic heating and cooling, drip irrigation with fertilizer and/or other chemicals injected into the water, trellis support for easy vertical training of the plants, automatic overhead curtains to keep the greenhouse from overheating during the day in the summer, and to keep the greenhouse warmer at night in the winter. Computer systems can provide efficient control of the greenhouse equipment and help provide optimum conditions for plant growth.

For seed harvesting and handling, it is useful to have a bulk seed extractor, washing screens, a seed sluice, and seed dryers. Seed companies have used such machines for years, and it is useful for the plant breeder to build smaller versions that match the size of the plant breeding program. Watermelon breeding is a labor intensive job, but mechanization can help make the most of the available workers, funds, and time.