基于分位数回归模型分析小麦千粒重与气候因子的关系  

Relationship between Thousand-Seed Weight of Wheat and Climate Factors According to Quantile Regression Model

作　　者：唐卷[1,2,3] 汤永禄[4] 李朝苏[4] 陈芋文 陆见光 向林泓 

机构地区：[1]中国科学院成都计算机应用研究所,四川成都610041 [2]中国科学院重庆绿色智能技术研究院,重庆401122 [3]中国科学院大学,北京100049 [4]四川省农业科学院作物研究所,四川成都610066

出　　处：《西南农业学报》2014年第3期943-949,共7页Southwest China Journal of Agricultural Sciences

基　　金：四川省科技创新苗子工程基金资助项目(2012ZZ084);国家科技支撑计划项目(2012BAD20B06);重庆市农情气象专家决策系统技术研发项目(cstc2012gg-yyjs0606);重庆市科技攻关计划项目(cstc2011ggB40027)

摘　　要：粒重作为禾谷类作物关键产量因子之一,其稳定性直接关系到产量在年际或环境间的稳定性。基于长期定位试验数据,采用普通最小二乘回归法(OLSR)和分位数回归法(QR)分析了小麦千粒重与气候因子之间的关系。结果表明:与普通最小二乘回归分析相比,分位数回归分析除了能揭示小麦千粒重与气候因子之间的一般规律外,还能细致地反映气候因子与千粒重尾部特征的关系;并能够动态地描述出小麦不同水平千粒重与各气候因子之间的变化规律。籽粒灌浆第一阶段和第三阶段的日平均温度、第二阶段的日照时数和降雨量随小麦千粒重变化呈一条倒"U"型曲线。与处于低粒重水平的QR分析结果相比,OLSR分析严重低估了第三阶段日平均温度和第二阶段降雨量的负影响;与处于高粒重水平的QR分析结果相比,OLSR分析严重低估了第一阶段日平均温度的负影响。Grain weight was one key factor of production of cereal crop, and its stability was directly related to the stability of production be- twecn the inter-annuals or environments. Based on the long-term experimental data, in this paper, the relationships between thousand-seed weight of wheat and climate factors were analyzed with ordinary least squares regression （OLSR） and quantile regression （QR） method. The result showed that compared with the OLSR analysis, the QR analysis not only revealed the general rules between thonsand-seod weight of wheat and climate factors, but also meticulously reflected the relationship between climate factors and thonsand-seed weight of upper or lower quantiles. And it dynamically described the change law between thousand-seed weight of different quantiles and climate factors. The change rule of daily average temperature on the first stage and on the third stage of grain filling, sunshine duration on the second stage and rainfall on the second stage of grain firing along with thousand-secd weight of different quantile were reverse ＇ U＇ curves. Compared with the result of the QR analysis in lower quantiles, the OLSR analysis seriously underestimated the negative influence of daily average temperature on the third stage and rainfall on the second stage. And oppositely, the OLSR analysis seriously underestimated the negative influence of daily aver- age temperature on the first stage.

关 键 词：小麦 气候因子 千粒重 普通最小二乘回归 分位数回归 

分 类 号：S512.1[农业科学—作物学]