基于投影寻踪模型的甘蓝型油菜产量因素评价

明确油菜产量因素对产量的贡献对于有目的地选择油菜优良组合具有重要意义，而常用的评价油菜产量因素贡献大小的分析方法限制较多。为此，本研究以正常生长的12份早熟和38份晚熟甘蓝型油菜组合为材料，利用基于加速遗传算法的投影寻踪模型分别定量评价了早、晚熟油菜组合产量因素对产量的贡献，并将其与灰色关联度分析、主成分分析结果进行了对比分析，讨论了各分析方法对产量因素贡献评价的准确性。结果表明，早熟组合产量贡献最大的3项指标依次为主序有效角果数 > 有效分枝部位 > 二次有效分枝数，贡献率分别达到36.79%、24.02%和11.33%；晚熟组合产量贡献最大的3项指标依次为每角粒数 > 千粒重 > 有效分枝部位，贡献率分别达到29.81%、17.52%和14.75%。有效分枝部位对不同熟期组合产量均有较大贡献；除此之外，早熟组合产量的形成还较多依赖于生育前期的有效分枝数与有效角果数，晚熟组合产量的形成则较多依赖于生育后期的每角粒数与千粒重。基于投影寻踪模型的油菜产量因素评价能较准确地判断油菜实际产量，并在利用各种分析方法评价油菜产量因素对产量的贡献中发现，灰色关联度分析和主成分分析比投影寻踪模型评价准确性低。从油菜产量因素评价及产量预测综合考虑，基于投影寻踪模型的油菜产量因素评价及产量预测具有最高的准确性，可以优先作为甘蓝型油菜良种选育及产量预测的一种分析方法。

Evaluation of factors affecting rape (swede type) yield using the projection pursuit model

The relative contributions of factors that affect rapeseed production were determined. A field experiment was conducted in Jianyang, Sichuan using 12 early-maturing and 38 late-maturing newly recombinant rapeseed varieties. Contributions were evaluated using the projection pursuit model based on real code deaccelerating genetic algorithm. The evaluation was compared with gray correlation analysis and principal component analysis. Effective pods in main inflorescence, effective branching position, and second effective branch number were the major contributors for early-maturing varieties, accounting for 36.79%, 24.02%, and 11.33% of the yield variation, respectively. For late-maturing varieties, seeds per silique, 1000-grain weight, and effective branching position were the most influential factors, accounting for 29.81%, 17.52%, and 14.75% of the yield variation, respectively. Interestingly, effective branching position was a significant contributor for both early- and late-maturing varieties. In addition, yield appeared to be influenced mostly by the number of effective branches and the number of effective pods, both formed during early growing stages, for early-maturing rape plants, and by seeds per silique and 1000-grain weight, formed during late growing stages, for late-maturing varieties. Predicted yields by the projection pursuit model were consistent with observed yields. Rapeseed yield was affected mostly by branching and pod formation for early-maturing varieties and by seed development for late-maturing varieties, and the yield potential was accurately predicted by the projection pursuit model.