Burdon R D. 1977. Genetic correlation as a concept for studying genotype-environment interaction in forest tree breeding. Silvae Genetica, 26(5–6): 168–175.
众所周知，基因型相对于彼此的表现可以根据环境而变化，使得在一个环境中优越的基因型在其他地方可能不会相应优于其他地方。这种现象已经在基因型环境相互作用的概念中得到了体现。
在育种计划中，需要确定相互作用的规模和模式，以便使遗传增益预测成为决策的辅助。近年来，许多论文出现在基因型环境相互作用上，其中许多论文一直关注基因型相互作用行为的统计学特征（如FINLAY和WILKIN SON，1963; PERKINS和JINKS，1968 a，b; BREESE ，1969; FREE MAN和PERKINS，1971; EBERHART和RUSSELL，1966; FREE MAN，1973; PEDERSON，1974;见SHELBOURNE，1972）。不过，考虑这个重点的情况是很好的。大部分工作都是每年必须收获的作物，如果每年必须收获，那么任何一个季节的增长和产量必须自己考虑。在每个地点，任何一年的天气和文化处理有效地产生了独特的环境，并引入了环境随机变化的重要因素。此外，有关作物通常只有少数或多或少的稳定品种被考虑。
在森林繁育方面，情况是非常不同的，在这里，有人认为，主要的注意力应该矛盾地被赋予环境的作用而不是基因型在产生互动中的作用。在某种程度上，这个原则已经在区域化树种育种计划的实践中已经是隐含的，但值得仔细研究这个原则的基础。森林环境主要是永久性的特征，因为我们不能改变气候，地形或许多土壤性质。年度气候波动的影响往往在旋转中平均化，这是一个大大消除作物环境变化的一个因素的因素。即使初期生存在一年内较差，因此一年的种植通常不会是砍伐计划的离散作物，因此影响减弱。相比之下，基因型不是永久性的特征，因为新的可以被创造整个时间，我们最终将需要考虑和筛选大量的它们。不是在全面的环境范围内测试所有这些基因型，而是在几个已被表征的环境中评估每个基因型将更容易，以便允许对其余环境的令人满意的外推性能。另一方面，树种繁殖的重点应该是适合于特定环境的生产基因型，或一方面适用于广泛环境的基因型，这一点是有争议的。然而，这两种对比的方法要求在相互作用方面对环境进行表征。与前者一样，需要划定适当的环境群体，并确定组内哪些特定环境能最佳地解决遗传差异。使用后一种方法，需要认识到一些环境，其单独或联合地为广泛适应的基因型提供有效的筛选。无论如何，通常需要从各种可能的选择中评估预期的遗传增益。我将首先研究在研究互动中传统使用差异分析，以纯粹的统计意义和表征环境的作用来说明其缺陷。然后，我将介绍替代概念，环境之间的遗传相关性，并涵盖其估计，一些应用和其一些统计属性。这种分析方法主要适用于遗传原料中没有明显的地理适应框架的外来物种的情况。
It is well known that the performance of genotypes relative to each other can vary according to the environment so that genotypes which are superior in one environmen may not be correspondingly superior elsewhere. This phenomenon has been formulated in the concept of genotype environment interaction.
In breeding programmes one needs to ascertain the magnitude and the pattern of interactions, to allow genetic gain prediction as an aid to decision making. During recent years numerous papers have appeared on genotype environment interactions, and many of these papers have been concerned with the statistical characterisation of the interactive behaviour of genotypes (e.g., FINLAY and WILKIN SON, 1963; PERKINS and JINKS, 1968 a, b; BREESE, 1969; FREE MAN and PERKINS, 1971; EBERHART and RUSSELL, 1966; FREE MAN, 1973; PEDERSON, 1974; see SHELBOURNE, 1972). It is well, though, to examine the circumstances underlying this emphasis. Much of the work was with crop plants which were annual or, if perennnial, had to be harvested annually, so that the growth and yield in any one season must be considered in its own right. At each location the weather and cultural treatment in any year effectively produce a unique environment, and introduce an important element of random variation among environments. Moreover, with the crops concerned there were often only a few more or less stabilised cultivars to be considered.
With forest tree breeding the situation is very different, and here it is contended that main attention should paradoxically be given to the role of environments rather than of genotypes in generating interactions. To some extent this principle is already implicit in the practice of regionalising tree breeding programmes, but it is worth taking a closer look at the basis for the principle. Forest environments are largely permanent features, since we can do little to change climate, topography, or many of the soil properties. The effects of year-to-year climatic fluctuations tend to be averaged out over the rotation, a factor which largely removes one element of variation in the crop en vironment. Even if initial survival is poor in one year the effects are damped down by the fact that one year’s planting is not normally a discrete crop with respect to felling plans. Genotypes, by contrast, are not permanent features in that new ones can be created the whole time, and we will eventually need to consider and screen large numbers of them. Rather than test all these genotypes over a comprehensive range of environments, it would be much easier to evaluate each genotype on just a few environments which are already characterised so as to permit satisfactory extrapolation of performance to the remaining environments. It is debatable whether the emphasis in tree breeding should be on producing genotypes suitable for specific environments, on one hand, or genotypes suited to a wide range of environments, on the other. Yet both these contrasting approaches demand the characterisation of environments with respect to interaction. With the former, one needs to delimit appropriate groups of environments, and to identify which particular environments within groups give best resolution of genetic differences. With the latter approach one needs to recognise a few environments which, singly or in conjunction, provide effective screening for genotypes which are widely adapted. In any case one will generally need to evaluate the expected genetic gains from the various possible options. I will first examine the traditional use of analysis of variance in studying interactions to illustrate its deficiencies both in a purely statistical sense and in characterising the role of environments. I will then introduce the alternative concept, genetic correlation between environments, and cover its estimation, some applications, and some of its statistical properties. This analytical approach is primarily suitable for cases such as exotic species in which there is no obvious framework of geographic adaptation within the genetic raw material.
