Genotype by environment interactions affecting grain sorghum. III. Temporal sequences and spatial patterns in the target population of environments

Abstract

The variable nature of rainfall in north-eastern Australia confounds the process of selecting sorghum hybrids that are broadly adapted. This paper uses a crop simulation model to characterise the drought environment types (ET) that occur in the target population of environments (TPE) for dryland sorghum. Seventy seasons (1921–1990) of simulations of the yield of a sorghum genotype and the associated within-season sequence of a stress index were conducted for a small TPE of 6 locations and also for a large TPE of 211 locations that attempted to represent the entire sorghum region.

Previously, using the small dataset of 6 locations, pattern analysis enabled us to group seasonal stress indices from each trial into major ETs: ‘low terminal stress’ (ET1), severe terminal stress (ET2), and intermediate mid-season/terminal stress (ET3) in the ratio 33 : 38 : 29. When the dataset was broken into a sequence of 16 multi-environment trials (METs), each of 3 years and 6 locations, the ratios of ET1 : ET2 : ET3 differed greatly among METs, i.e. any single MET was not randomly sampling the TPE. Hence, for any MET, the average yield (GV_u) was not the same as the overall mean of the entire 70-year dataset. If the trial yields were weighted according to the ratio of ET1 : ET2 : ET3 in the overall TPE, then GV_w (s.d. = 0.13) for a single MET was much closer to the overall mean than was GV_u (0.38). For different METs, the values of GV_w were up to 30% higher or 15% lower than GV_u. Across METs, the difference between GV_u and GV_w was positively correlated (r = 0.88, n = 16, P < 0.05) with the frequency of ET1 (‘low terminal stress’) encountered within the MET and negatively correlated (r = −0.82) with the frequency of ET2. The value of weighting was confirmed by its ability to verify that two simulated genotypes had the same mean yield over many trials, even though they differed in their specific adaptation to the different ETs.

The large TPE consisted of more than 15 000 simulations and was classified in 2 stages (within/among locations), repeated for each of 3 soil types. In years in which the simulation sowing criteria were met, the ratios of ET1 : ET2 : ET3 were about 4:2:4, 4:5:1, and 6:3:1 in the shallow, intermediate, and deep soils, respectively. Hence, over all soil types and locations, the sorghum TPE for northern Australia consists of at least 30% each of low terminal stress (ET1) or severe terminal stress (ET2) and these environment types need to be sampled. The incidence and nature of the ‘intermediate midseason/terminal stress’ environment type (ET3) varies with soil type and location.

Weighting genotype performance should improve the precision of the estimate of its broadly adapted value, and be of practical use in breeding programs in these variable environments. Although the ‘boundary conditions’ of the TPE are not yet resolved, this paper also shows that simulation and pattern analyses can be used to determine the structure of the abiotic TPE. Taking other factors into account (e.g. soil type distribution, shire production levels, and farm profit), selection trials could be weighted to improve selection for narrow or broad adaptation, depending on the purpose of the breeding program.