Efficient Trialing Methods for Watermelon (Citrullus lanatus (Thunb.) Matsum. & Nakai)

Abstract

Researchers interested in evaluating watermelon (Citrullus lanatus (Thunb.) Matsum. & Nakai) cultivars for yield use multiple-row plots to simulate the monoculture system growers use, or single-row plots to save on land, labor, and seeds. We were interested in whether there is a significant interaction of border with center row when diverse cultivars are planted in adjacent rows. Charleston Gray, Crimson Sweet, and Sugar Baby were chosen to represent long, medium, and short vined cultivars, respectively. Cultivars were planted in three-row plots with all nine combinations of the three represented in border and center rows. Each cultivar combination of center row and border rows represented one treatment. The experiment was a randomized complete block with nine plot treatments, two locations (Kinston, Clinton), and three replications. Vine length was measured during the season, and fruit were graded (marketable and cull), counted and weighed at four harvests. Results showed that Charleston Gray had the longest vines, followed by Crimson Sweet and Sugar Baby. In the analysis of variance, the largest effects (F ratio size) on yield were from cultivar, location, and the interaction of the two. The smallest effects were due to the interaction of center with border row, although center by border interactions were significant (5% level) in some cases. Therefore, researchers interested in running trials with many cultivars and small seed quantities can obtain good data using single-row plots. However, there is a small (but significant) interaction of center with border in some cases, so testing at the final stage should be with trials having multiple-row plots or grouping cultivars by vine length. Cultivars having extreme plant types (dwarf vines for example) should be tested in separate trials. One of the most expensive stages of breeding is field testing. This encourages breeders and researchers to make efficient use of limited land, labor and seed in order to maximize information obtained while minimizing the costs of trialing. We were interested in whether smaller single-row plots could be used that would effectively be able to achieve and predict yields obtained from plots with larger dimensions when diverse cultivars were evaluated. The 13 cultivars Allsweet, Fiesta, Regency, Starbrite, Sultan, Florida Favorite, Charleston Gray, Hopi Red Flesh, Crimson Sweet, Jubilee, Navajo Sweet, New Hampshire Midget, and Sugar Baby were used to represent a wide range in yield. Cultivars were planted in single-row plots of three different plot lengths (7.3 m, 3.7 m, and 2.4 m) with all cultivars represented in each plot size. Each combination of cultivar and plot size represented one treatment. The experiment was a randomized complete block with 39 plot treatments, two locations (Kinston and Clinton) and three replications. Fruit were graded (marketable and cull), counted and weighed at five harvests. Analysis of variance indicated the largest effects (F ratio size) on plot yields were from location, plot size and cultivar. The smallest effects were due to the interactions of location with plot size, and cultivar by location by plot size. A location by cultivar interaction was also present, but F ratios were small indicating a small effect. Yields from 7.3 m and 3.7 m plots were consistently no different from each other. Regression analysis of 2.4 m and 3.7 m plots in prediction of 7.3 m plot yields showed 3.7 m plots to have higher R2 (0.90), lower mean square error, lower standard deviations and a lower coefficient of variation than was found for 2.4 m plots. Therefore, researchers interested in maximizing information obtained while minimizing costs in trials with many cultivars can obtain data representative of large 7.3 m plots using 3.7 m plots. However, alleys separating 7.3 m and 3.7 m plots required a yield correction to compensate for the extra growing space allotted to these plots.