基于多元统计分析的冬小麦干旱综合指标构建及监测研究  

Construction and Monitoring of Comprehensive Indicators of Winter Wheat Drought Based on Multivariate Statistical Analysis

作　　者：谢永凯 冯美臣[2] 秦明星[3] 杨武德[2] 刘敏 孟万忠 XIE Yongkai;FENG Meichen;QIN Mingxing;YANG Wude;LIU Min;MENG Wanzhong(Institute of Geography Science,Taiyuan Normal University,Jinzhong,Shanxi 030619,China;College of Agriculture,Shanxi Agricultural University,Taigu,Shanxi 030801,China;College of Resources and Environment,Shanxi Agricultural University,Taigu,Shanxi 030801,China)

机构地区：[1]太原师范学院地理科学学院,山西晋中030619 [2]山西农业大学农学院,山西太谷030801 [3]山西农业大学资源环境学院,山西太谷030801

出　　处：《天津农业科学》2023年第5期17-24,共8页Tianjin Agricultural Sciences

基　　金：山西省基础研究计划项目(202203021212188,20210302123411);山西省高等学校科技创新项目(2021L444);太原师范学院大学生创新创业训练项目(CXCY2208);国家自然科学基金项目(31871571)。

摘　　要：为了实现水分胁迫后冬小麦干旱指标综合表现的定量监测,以2017—2018、2018—2019年的冬小麦水分胁迫试验为基础,选择冬小麦叶片含水量(LWC)、叶绿素密度(ChD)、游离脯氨酸含量(Pro)以及抗氧化物酶中的超氧化物歧化(SOD)、过氧化氢酶(CAT)和过氧化物酶(POD)活性等生理参数作为研究对象,利用主成分分析方法(PCA)构建了冬小麦干旱综合指标(Comprehensive drought index,CDI)。结合相关分析法和逐步多元线性回归(CA+SMLR)、偏最小二乘法和逐步多元线性回归(PLS+SMLR)及连续投影算法(SPA)对光谱反射率进行了特征波段提取,综合利用化学计量学方法,对冬小麦生理生化及CDI指标监测展开了研究。结果表明:通过CA+SMLR提取的特征波段个数较少,并且所构建的SMLR模型表现一般;利用SPA构建的监测模型表现优于CA+SMLR和PLS+SMLR 2种方法,可以实现对冬小麦CDI指标优化目的。利用多元回归分析方法构建的模型对比,发现基于全谱建立的PLSR模型表现(R^(2)=0.885,RMSEC=0.221,RPD=2.772;R^(2)=0.631,RMSEP=0.441,RPD=1.625),其预测效果最好;SPA方法提取特征波段建立的MLR模型表现(R^(2)=0.647,RMSEC=0.387,RPD=1.355;R^(2)=0.672,RMSEP=0.376,RPD=1.500)次之。综上,通过CDI模型的构建,为实现水分胁迫后冬小麦生理参数综合表现的高光谱监测提供了参考。The study was based on tests of winter wheat after water stress that from 2017—2018 and 2018—2019 to realize the quantitative monitoring of drought index.By selecting physiological parameters such as leaf water content(LWC),chlorophyll density(ChD),free proline content(Pro)and superoxide dismutase(SOD),catalase(CAT)and peroxidase(POD)activities of antioxidant enzymes of winter wheat as a comprehensive drought index(CDI)for winter wheat was constructed using principal component analysis(PCA).Combined with correlation analysis and stepwise multiple linear regression(CA+SMLR),partial least square method and stepwise multiple linear regression(PLS+SMLR)and successive projections algorithm(SPA),the important band extraction was carried out,comprehensive use of stoichiometry methods,physiological and biochemical and CDI monitoring of winter wheat were studied.The results showed that the number of important bands extracted by CA+SMLR was low and the performance of the SMLR model was average.The model constructed using SPA outperformed both CA+SMLR and PLS+SMLR,and could achieve the purpose of optimizing the CDI indicators for winter wheat.Comparing the models constructed using multiple regression analysis methods,it was found that the performance of the PLSR model built based on the full spectrum(R^(2)=0.885,RMSEC=0.221,RPD=2.772;R^(2)=0.631,RMSEP=0.441,RPD=1.625)was best performance,and the performance of the MLR model built by the SPA method of extracting the important bands(R^(2)=0.647,RMSEC=0.387,RPD=1.355;R^(2)=0.672,RMSEP=0.376,RPD=1.500)was the second performance.The construction of the CDI model provides a reference for achieving hyperspectral monitoring of the integrated performance of physiological parameters of winter wheat after water stress.

关 键 词：主成分分析 干旱综合指标 特征波段 模型表现 

分 类 号：S512.11[农业科学—作物学]