Noncontiguous replications in soybean performance tests

Soybean [Glycine max (L,) Merr,] performance trials are commonly conducted in randomized complete block designs with contiguous replications. Local environmental hazards such as hail, flooding, or herbicide drift can result in the loss of entire experiments and the included germplasm with contiguous replications. Noncontiguous replications, spatially separated at a location, minimize chances of local hazards affecting all replications. Our objective was to compare the validity of data from noncontiguous vs. contiguous replications of soybean performance trials. Eleven performance tests were conducted over a 3-yr period in three-replicate randomized complete block designs using both contiguous and noncontiguous replications. Distance between noncontiguous replications varied from 200 to 800 m. Twenty-five to 40 lines in either Maturity Group II or m were evaluated in each test. Traits measured were plant maturity, mature plant height, and seed Yield. Variability among lines was significant (P = 0.01) using both contiguous and noncontiguous replications in all 11 tests for plant maturity and plant height. Variability among lines was significant (P = 0.05) for seed yield in 7 of 11 tests of contiguous replications and in 9 of 11 tests of noncontiguous replications. Correlation coefficients among line means between the two types of replication varied from 0.83 to 0.98 for maturity, from 0.51 to 0.94 for plant height, and from 0.51 to 0.89 for seed yield in those tests where significant variation was observed among lines. Evaluating soybean lines in tests using noncontiguous replications requires no more effort or spare than using contiguous replications. The results demonstrate that noncontiguous replications are an effective and useful alternative to contiguous replications for soybean performance tests.