条锈病胁迫下冬小麦冠层叶片氮素含量的高光谱估测模型

Hyperspectral estimation models for LTN content of winter wheat canopy under stripe rust stress

为利用高光谱遥感诊断条锈病胁迫下作物的营养状况,测量感染条锈病的冬小麦冠层反射率以及相应叶片全氮(LTN)含量,利用线性和非线性回归方法,建立了微分光谱与小麦LTN含量之间的回归模型。研究表明随病情加重,小麦LTN含量逐渐降低,并与一阶微分光谱在430～518、534～608、660～762 nm以及783～893 nm区域具有极显著相关性。经检验,以红边内一阶微分总和与蓝边内一阶微分总和比值(SDr/SDb)为变量的模型是估测LTN含量的最佳模型,其RMSE为0.3567,相对误差为8.33%。因此,利用高光谱遥感估测条锈病胁迫下作物LTN含量是可行的,且具有较高的反演精度。研究成果可为小麦氮素营养监测、精准施肥以及条锈病情诊断等提供理论依据和指导。

The objective of this study is to diagnose the nutrition status of crops by hyperspectral data under stripe rust stress. Canopy reflectance of winter wheat infected by stripe rust was measured in situ, and the leaf total nitrogen （LTN） contents corresponding to the spectra were determined in laboratory. Linear and non-linear regression methods were used to build the regression models between derivative variables and LTN content. It is shown that LTN of disease wheat gradually decreases with disease aggravating and there is high correlation between LTN and first derivative data at 430- 518, 534-608, 660-762 nm and 783-893 nm By validation, the model consisting of the ratio of sum of the first derivative within red edge and sum of the first derivative within blue edge （SDr/SDb） had the best performance, and the RMSE was 0.3567 and the relative error was 8.33 %. So it is feasible to estimate LTN content of crops under disease stress by those proposed models based on hyperspectral remote sensing and the accuracy is satisfactory. These results also provide a theoretical basis for monitoring of plant nitrogen status and for diagnosing disease severity of wheat stripe rust and precision management of nitrogen fertilization in wheat production.