Crop Performance Trials under Irrigated and Dryland Conditions

Abstract

Corn (Zea mays L.) and soybean [Glycine nuJX (L.) Men·.] perfonnance trials are sometimes conducted under both irrigated und dryland conditions in the southeastern USA. It is not known whether irrigated trials are better trials statistically or if irrigation causes l"l~lative rank changes in hybrid or cultivar perfonnance. To answer these questions, corn data were examined from four different environmemts and three maturity groups, and soybean data were examined from five different environments and three maturity groups between 1987 and 1990. Water management x hybrid/cultivar interactions were evident in 8 of 12 corn data sets and 7 of 16 soybean data sets. Combining data across years eliminated the water management x hybrid/cultivar interac· tions in all soybean data sets and five of nine corn data sets. Irrigated corn trials produced year x hybrid interactions in seven ·~f nine data sets while dryland trials produced year x hybrid interactions in four of nine data sets. Irrigated trials were only slightly better than dryland trials in accounting for variation; i.e., larger r values but produced smaller error variances in four of 12 data sets of corn. Irrigated trials separated entries better in nearly half of the soybean data sets. The data suggested that irrigated trials were not statistically llletter trials than dryland trials in all situations. Water management x hybrid/ cultivar interactions suggest that data for irrigated and d1yland trials should be reported separately, and the year x hybrid/cultivar interactions indicated that multi-year data should be examine~! during hybrid/cultivar selection. Error variances, F probability values, and r values should be examined to detennine level of precision when comparing trials of equal composition.

Measures of Validity in Cultivar Performance Trials

“…However, until the regression values approach 13 = 2.0, the probability of CV values less than some desired level increases with the mean, suggesting that those who conduct performance trials would be better off conducting their trials to achieve maximum yields. For example, Bowman et al (1993) showed that irrigation per se did not decrease error variances inz ' creaseR values, or change entry F-tests; CVs were not used for comparison in that study, but were lower in the irrigated trials than the dryland trials (8.5 vs. 14.2, respectively, in corn, 10.7 vs. 14.0, respectively, in soybean).…”

“…Other measures of precision or relative variation include Gini's coefficient of concentration (Kendall and Stuart, 1977), error mean square (EMS), and coefficient of determination (R 2 ) (Bowman et al, 1993;Askew et al, 1994). As a measure of precision, EMS is valid only for comparison of similar trials measured in the same units and thus would not be a useful statistic to report in cultivar performance data, but would be useful to the researcher in determining the validity of a trial (Bowman and Rawlings, 1995).…”