A comparison of statistical methods to describe genotype x environment interaction and yield stability in multi-location maize trials
Loading...
Date
2004
Authors
Alberts, Martin J. A.
Journal Title
Journal ISSN
Volume Title
Publisher
University of the Free State
Abstract
The objectives of the study were to evaluate different statistical methods to describe genotype x environment interaction over three years, with maize hybrids across several environments. The environment and soil variables have a major effect on the performance of maize hybrids in South Africa. The second objective was to evaluate and compare the different statistical stability models and procedures, to identify the best stability model to accurately assess and rank maize hybrids according to their stability over environments and years. The third objective was to describe genotype x environment interaction and the adaptation of maize hybrids to different environments over years. Twenty three maize hybrids were evaluated at 42 environments between 2001 and 2003 in the major maize producing areas in South Africa. Grain yield and other agronomic traits were determined but mean grain yield was used to determine stability with the following stability procedures: CV (coefficient of variation), Linn and Binns, Shukla, Wricke, Finlay and Wilkinson, Eberhart and Russell, and the ASV (AMMI stability value). The comparison of the procedures were done with Spearman’s rank correlation coefficient and the significance determined with student’s t-test. Linn and Binns cultivar performance measure ranked the hybrids, with high (Pi) values as the most stable. CRN 80-10 was ranked first, CRN 3505 ranked second and SNK 8520 ranked third. The unstable hybrids with low (Pi) values were SNK 6025, PHB 32A03 and PAN 6615. Linn and Binns procedure was not significantly correlated, with any of the other procedures. It was only significantly correlated with mean yield (r = 0.97332**), thus confirmed that it is more a measure of performance and not really a stability parameter. Finlay and Wilkinson’s regression coefficient indicated that DKC 80-10, CRN 3505 and PAN 6568 showed average stability and were adapted to most of the environments. SB 7551, PHB 3203W and SA 7401 have below average stability and adapted to the higher yielding environments. DK 617, DKC 61-24 and SNK 8520 were of average stability but were specifically adapted to low yielding environments. This method was also not comparable to the other methods and was only positive and significantly correlated with CV. Shukla’s stability variance indicated that CRN 3549, PAN 6615, DKC 63-20, PAN 6573 and SA 7401 were stable and SNK 2551, CRN 4760B, CRN 3505, PHB 3203W and SNK 8520 were unstable hybrids. DKC 80-10 was the highest yielding hybrid but only average on stability. This method compared well with the procedures of Eberhart and Russell, Nassar and Hühn, Wricke and the ASV (AMMI). The comparison of the rank correlations were all significant and positive. Shukla’s and Wricke’s methods had total correspondence (r = 1.000**). These methods will rank hybrids equivalently according to their stability. Wricke’s ecovalence ranked CRN 3549, PAN 6615, DKC 63-20 and PAN 6573 as the most stable hybrids with SNK 2551, CRN 4760B, CRN 3505, and PHB 3203W as the most unstable hybrids. Wricke’s ecovalence was positively and significantly correlated with Shukla, Eberhart and Russell, Nassar and Hühn and ASV. Eberhart and Russell’s deviation from regression indicated that CRN 3549, PAN 6615, SA 7401, DKC 63-20 and PAN 6573 were the stable hybrids. SNK 2551, CRN 4760B and CRN 3505 were the unstable hybrids. It corresponded with the methods of Shukla, Wricke, Nassar and Hühn and the ASV. Nassar and Hühn’s mean absolute rank method indicated PAN 6615, CRN 3549, SNK 6726 and DKC 80-10 as the stable hybrids. CRN 3505, CRN 760B, SNK 2551 and DKC63-20 were the most unstable hybrids. This method was significantly and positively correlated with Shukla, Wricke and Eberhart and Russell. It was not correlated with ASV, Linn and Binns and Finlay and Wilkinson. The AMMI stability value (ASV) ranked DKC 63-20, DK 617, DKC 61-24 en SB 7551 as stable. SNK 2551, CRN 4760B, PHB 3203W en CRN 3760 was unstable. ASV was significantly correlated with the methods of Shukla, Wricke and Eberhart and Russell. Die AMMI model 2 indicated Delmas (quadrant II), Petit, Meerlus, Bergville, Piet Retief and Ermelo (quadrant III) as the high yielding environments. DKC 80-10, CRN 4760B, PAN 6568, SNK 8520 and SA 7401 were adapted to the high yielding environments but specific to Delmas. CRN 3505, SB 7551, SNK 2551 and CRN 3549 are also adapted to high yielding environments but more specific to Petit, Meerlus, Bergville and Ermelo. Kroonstad, Kameel, Bothaville, Viljoenskroon and Rietgat are the average to low yielding environments and are clustered in quadrant I. DK 617, SNK 6726, DKC 63-20 and DKC 61-24 were adapted to these environments. Ficksburg, Wonderfontein and Ogies were clustered in quadrant IV and shown to be average to low yielding environments, hybrids that were specific adapted to these environments were PHB 3203W, PHB 32A03, DKC 71-21, PAN 6615 and PAN 6573. Die AMMI method successfully summarized patterns and relationship of the hybrids with the environments. AMMI indicated the genotype x environment interactions and clustered the hybrids according their adaptability to certain environments. The graphical AMMI biplot explained and described the hybrid’s adaptation and interaction with the environments.
Description
Keywords
Genotype-environment interaction -- Statistical methods, Corn -- Genetics, Corn -- Yields -- Statistical methods, Dissertation (M.Sc.Agric. (Plant Sciences (Plant Breeding))--University of the Free State, 2004