基于连续小波变换估测干旱胁迫下玉米籽粒产量  

Estimation of maize grain yield under drought stress based on continuous wavelet transform

作　　者：邹佳琪 王仲林 谭先明 陈燎原 杨文钰[1] 杨峰[1] ZOU Jia-Qi;WANG Zhong-Lin;TAN Xian-Ming;CHEN Liao-Yuan;YANG Wen-Yu;YANG Feng(College of Agronomy,Sichuan Agricultural University/Key Laboratory of Crop Ecophysiology and Farming System in Southwest,Ministry of Agriculture and Rural Affairs/Sichuan Engineering Research Center for Crop Strip Intercropping System,Chengdu 611130,Sichuan,China;Rice Research Institute,Sichuan Agricultural University,Chengdu 611130,Sichuan,China)

机构地区：[1]四川农业大学农学院/农业农村部西南作物生理生态与耕作重点实验室/四川省作物带状复合种植工程技术研究中心,四川成都611130 [2]四川农业大学水稻研究所,四川成都611130

出　　处：《作物学报》2024年第4期1030-1042,共13页Acta Agronomica Sinica

基　　金：国家重点研发计划项目(2022YFD2300902)资助。

摘　　要：利用高光谱遥感技术监测作物水分状况和籽粒产量,对于调控作物生长、优化水分管理和改善产量形成具有重要意义。本研究玉米品种选用正红505,于2018—2019年在四川雅安和仁寿的试验田设置4个水分处理(正常水分、轻度、中度和重度干旱),分析玉米在拔节期(V6)、抽雄期(VT)和灌浆期(R^(2))的冠层含水量(canopy water content,CWC)与籽粒产量的定量关系,利用植被指数和连续小波变换对光谱反射率数据进行处理,采用线性回归方法构建CWC定量反演模型,进一步探索以CWC为桥梁建立的玉米籽粒产量的预测模型效果。结果表明,(1)利用小波特征构建的CWC估测模型的预测效果高于植被指数,V6、VT和R^(2)期分别以小波特征gaus3770,64、rbio3.31635,2和rbio3.3838,2构建的线性回归模型检验精度较高,R^(2)分别为0.770、0.291和0.233。(2)CWC与玉米籽粒产量间建立的线性回归模型均达极显著水平(P<0.01),V6、VT和R^(2)期的R^(2)分别为0.596、0.366和0.439。(3)基于光谱反射率构建的产量预测模型以V6期小波特征gaus3770,64的验证效果最好(R^(2)=0.577,RMSE=1.625 t hm^(–2)),可作为预测玉米籽粒产量的最佳时期。因此,本研究提出的“光谱反射率—冠层含水量—产量”建模方法能够实现对玉米籽粒产量的精确估测,为未来大面积监测玉米生产力提供了理论依据。The use of hyperspectral remote sensing technology to monitor crop water status and grain yield is important for regulating crop growth,optimizing water management and improving yield formation.Zhenghong 505 was selected as the maize variety in this study,to analyze the quantitative relationship between canopy water content(CWC)and grain yield of maize at jointing stage(V6),tasseling stage(VT),and filling stage(R2),four drought stress treatments(well-watered,mild,intermediate and severe drought)were conducted in the experimental fields of Ya’an and Renshou in Sichuan Province from 2018 to 2019.The spectral reflectance data were processed using vegetation indices and continuous wavelet transform,and a linear regression method was used to construct a quantitative CWC inversion model to explore the effectiveness of CWC as a bridge to establish a spectral inversion model for maize grain yield estimation.The results showed that the CWC estimation models using wavelet features was better than that of vegetation indices,and the linear regression models constructed with wavelet features gaus3770,64,rbio3.31635,2 and rbio3.3838,2 at the V6,VT,and R2 stages had high test accuracy with the R2 of 0.770,0.291,and 0.233,respectively.The linear regression models established between CWC and maize grain yield all reached highly significant levels(P<0.01),with R2 of 0.596,0.366 and 0.439 at the V6,VT,and R2 stages,respectively.The yield prediction model based on the basis of spectral reflectance was the best validated with the wavelet feature gaus3770,64(R2=0.577,RMSE=1.625 t hm^(–2))at V6 stage,which can be used as the best period for predicting maize grain yield.Therefore,the“spectral reflectance-canopy water content-yield”modeling method proposed in this study can achieve an accurate estimation of maize grain yield and provide a theoretical basis for future large-scale monitoring of maize productivity.

关 键 词：玉米 籽粒产量 冠层含水量 植被指数 小波特征 

分 类 号：S513[农业科学—作物学]