Answers are grounded in this paper and cite their sources.
Multi-environment trials (METs) are routinely conducted as factorial combinations of genotypes, years, and locations but are usually analyzed with year–location combinations treated as single environments, which do not separate temporal, spatial, and spatiotemporal sources of genotype × environment interaction. Ten winter triticale (×Triticosecale Wittmack) cultivars were evaluated over three years (2019/2020–2021/2022) and at three locations in Bulgaria. We integrated factorial decomposition of genotype × year (G × Y), genotype × location (G × L), and genotype × year × location (G × Y × L) interactions with singular value decomposition of each component and canonical polyadic (CP/PARAFAC) decomposition of the three-way tensor (FITS framework). Location accounted for the largest share of total variation (36.59%), reflecting differences in mean productivity. Among genotype-related interactions, G × L (6.91%) exceeded G × Y (2.82%). The three-way interaction (4.18%) had a variance component as large as that of G × L. CP/PARAFAC re-expressed the unfolded SVD solution (84.41% of G × Y × L). Kasyno was the most interactive cultivar, and its three-way interaction was concentrated in one season. Among cultivars with small interaction, only Doni 52 also had above-average yield. In cross-validation, only the leading G × L axis improved prediction for untested year–location combinations. FITS provides descriptive information complementary to classical AMMI, which remains the basis for cultivar recommendation; its breeding applications require validation with larger datasets.
No takes yet. Share an insight, caveat, or question.
Stoyanov et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: