基于SIF-PLS模型的冬小麦条锈病早期光谱探测

为实现冬小麦条锈病早期探测、提高冬小麦产量和品质,研究了日光诱导叶绿素荧光(Solar-induced chlorophyll fluorescence,SIF)对冬小麦条锈病早期探测的可行性。基于3波段夫琅和费暗线(3-band Fraunhofer line discrimination,3FLD)和反射率荧光指数2种方法提取了冠层SIF数据,计算了对小麦条锈病敏感的光谱指数(Spectral index,SI),通过相关性分析优选了遥感探测小麦条锈病早期的特征参量,利用偏最小二乘(Partial least squares,PLS)算法构建冬小麦条锈病早期光谱探测模型。研究结果表明:O2-A波段的荧光强度(SIF-A)以及反射率荧光指数ρ440/ρ690、ρ675ρ690/ρ2683、ρ690/ρ655、ρ690/ρ600、DλP/D744、D705/D722均与小麦条锈病早期病情指数(Disease index,DI)达到了极显著相关,相关系数分别为-0.793、-0.523、-0.539、-0.497、0.541、0.446、0.490,可作为冬小麦条锈病早期光谱探测的荧光特征参量;基于3组SIF数据构建的PLS-SIF检验模型的决定系数分别为0.801、0.772、0.807,均方根误差分别为3.3%、3.1%、3.2%,较反射率光谱指数构建的SI-PLS模型决定系数至少提高了27%,均方根误差至少减少了24%。因此,冠层SIF数据更适于冬小麦条锈病的早期探测。本研究结果对及时进行冬小麦条锈病防控具有重要应用价值,可为利用卫星荧光遥感数据对小麦条锈病早期大面积、无损探测提供参考依据。

Early Detection of Winter Wheat Stripe Rust Based on SIF-PLS Model

Stripe rust is one of the main diseases that affects the production of winter wheat in China.The disease information was detected early in the winter wheat infection,and it is of great significance to prevent and control the disease and improve the yield and quality of winter wheat.The reflectance spectrum can reflect the change of concentration information of vegetation biochemical components,but it is greatly affected by the background noise,while the canopy solar-induced chlorophyll fluorescence(SIF)is less affected by the background noise and has certain photosynthetic physiological diagnosis capabilities.In order to study the feasibility of early detection of winter wheat stripe rust by SIF,the canopy SIF data was extracted based on two methods:3-band Fraunhofer line discrimination(3FLD)and reflectance fluorescence index.In order to explore the advantages of SIF in the early detection of wheat stripe rust,some SI sensitive to wheat stripe rust were obtained for comparison.The sensitivity of SIF and SI to wheat stripe rust early disease index(DI)was analyzed through correlation,and then the sensitive SIF and SI were used to construct the early wheat stripe rust spectrum detection model based on the partial least squares(PLS).The results showed that the fluorescence index SIF-A,ρ440/ρ690,ρ675ρ690/ρ2683,ρ690/ρ655,ρ690/ρ600,DλP/D744,D705/D722 extracted based on the radiance and reflectance method all had very significant correlation to the severity of wheat stripe rust,the correlation coefficients were-0.793,-0.523,-0.539,-0.497,0.541,0.446 and 0.490,respectively,which can be used as the chlorophyll fluorescence characteristic parameters for detection of winter wheat stripe rust.Based on the three sets of data,the determination coefficients of the PLS-SIF test model were 0.801,0.772 and 0.807,respectively,and the root mean square errors were 3.3%,3.1%and 3.2%,which were 27%at least higher than that of the SI-PLS model determination coefficients.The error was reduced by at least 24%.Therefore,canopy SIF data was more suitable for early detection of the severity of winter wheat stripe rust.The research results had important application value for timely prevention and control of winter wheat stripe rust,and provided a reference for the use of satellite fluorescence data for large-area,non-destructive detection of wheat stripe rust in the early stage.