许乃银, 张国伟, 李健, 等. 基于GGE双标图和比强度选择的棉花品种生态区划分. 中国生态农业学报, 2012, 20(11): 1500-1507.

Environmental interactions with genotype have heightened among plant breeders and/or extension agronomists the exploration and application of useful traits in the discrimination of genetic differences among candidate genotypes and similarities to average performances in target regions. GGE biplot with "which-wins-where" view has been a useful technique in graphical display and visualization of interrelation between environments and genotypes, and of mega-environment investigations. However, first two principal component restrictions in GGE biplots have led to uncertainties in terms of fitting degree optimizations of GGE models. Therefore a GGE biplot analysis was adopted in exploring and investigating possible mega-environments in a target cotton planting region. The region included fifteen test sites (Anqing, Nanyang, Huanggang, Jingzhou, Wuhan, Xiangfan, Changde, Yueyang, Nanjing, Nantong, Yancheng, Jiujiang, Jianyang, Shehong and Cixi) for cotton fiber strength selection in 27 independent test sets of cotton variety trials in the Yangtze River Valley for the period 2000-2010. The information ratio (IR) adjustment method was used as a validation tool to ensure that sufficient and necessary principal component scores were selected in the GGE biplot analysis. The study showed that the average effective fitting degree of the GGE biplot analysis of cotton fiber strength selection was 68.4%, with 13 exception records of over-fit or lost-fit. The counterpart fitting degreet of the IR-adjusted GGE model was 73.7%, with 6.0% improvement. This showed that optimization and adjustments to the GGE biplot using IR obviously enhanced the reliability of mega-environment investigations. Based on the GGE biplot analysis, the cotton planting region in the Yangtze River Valley was subdivided into four mega-environments for fiber strength selection � the first mega-environment included Anqing, Xiangfan, Nantong and Yueyang; the second comprised of Changde, Jiujiang and Wuhan; the third included Cixi, Nanjing, Huanggang, Jingzhou and Yancheng; and the forth also included Nanyang, Jianyang and Shehong. Based on the IR-adjusted GGE model, however, the cotton planting region more distinctively consisted of three mega-environments � the main mega-environment contained eleven out of fifteen test sites (Anqing, Wuhan, Jiujiang, Xiangfan, Nanyang, Yueyang, Changde, Huanggang, Jingzhou, Nanjing and Cixi), the minor mega-environment consisted of two test sites (Nantong and Jianyang), and the other minor mega-environment also consisted of two test sites (Yancheng and Shehong). Compared with the results of GGE biplot mega-environment analysis, IR-adjusted results were more reliable with distinct geographical characteristics for cotton fiber strength selection. Cotton plants in the Sichuan Basin and Jiangsu Coastal Regions were not suitable for fiber strength selection. It was therefore recommended to target the whole cotton planting region in the Yangtze River Valley. On the basis of the mega-environment investigation and above divisions, scientific decision-makings on cotton fiber strength selection, new cultivar breeding and policy registration for the Yangtze River Valley were recommended.

由于农作物品种区域试验是品种审定和推广应用的前提, 而区域试验中基因型与环境的互作效应是普遍存在的, 因而探索和利用试验环境在鉴别基因型遗传差异和对品种在目标环境中的平均表现代表性方面的作用以辅助品种选育和推广的问题, 越来越受到植物育种家和农技推广人员的高度关注。我们采用GGE双标图分析方法对2000-2010年期间27组长江流域国家级棉花品种区域试验的目标环境中可能存在的基于棉纤维比强度选择的品种生态区进行探索与划分, 并对品种生态区划分结果进行信息比(IR)校正, 以提高品种生态区划分的可靠性。结果表明: (1)基于纤维比强度选择的GGE双标图分析的总体有效拟合度为68.4%, 其中有13次出现过度拟合或拟合不充分现象, 总体拟合可靠性一般。而基于IR-GGE模型的总体拟合度为73.7%, 比GGE双标图的有效拟合度提高6.0%, 说明采用IR对双标图分析的结果进行优化和校正可以提高品种生态区划分的可靠性。(2)根据GGE双标图分析结果, 我国长江流域棉区大致可划分为4个基于纤维比强度选择的品种生态区, 第1个品种生态区包括安庆、襄樊、南通和岳阳, 第2个品种生态区包括常德、九江和武汉, 第3个品种生态区包括慈溪、南京、黄冈、荆州和盐城, 第4个品种生态区包括南阳、简阳和射洪。而基于IR较正的GGE模型则可划分为3个品种生态区: 第1个为主体品种生态区, 包括安庆、武汉、九江、襄樊、南阳、岳阳、常德、黄冈、荆州、南京和慈溪11个试验点, 第2个品种生态区包括南通和简阳, 第3个品种生态区包括盐城和射洪。IR校正后长江流域棉区的品种生态区划分更准确可靠, 地理区域特性也更明显, 说明地理环境因素对纤维比强度的选择效果仍然有很大的影响力, 四川盆地棉区和江苏沿海棉区并不适宜开展针对整个长江流域棉区的广适性棉花纤维比强度选择, 从而为长江流域棉区棉花纤维比强度的选择和推荐策略提供了科学的决策依据。