引用本文:
【打印本页】   【下载PDF全文】   查看/发表评论  下载PDF阅读器  关闭
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 178次   下载 337 本文二维码信息
码上扫一扫!
分享到: 微信 更多
协同冠层SIF和PRI光谱指数的构建及其在小麦条锈病监测中的应用
闫菊梅, 竞 霞, 张 腾,等
西安科技大学 测绘科学与技术学院
摘要:
【目的】利用反射率光谱在作物生物物理方面的优势和日光诱导叶绿素荧光(solar-induced chlorophyll fluorescence,SIF)、光化学反射率指数(photochemical reflectance index,PRI)在光合生理方面的优势,构建协同冠层SIF和PRI光谱指数(synergistic spectral index of SIF and PRI,SISP),旨在提高作物病害遥感探测精度。【方法】基于3FLD(three bands fraunhofer line discrimination)算法,估测小麦条锈病在不同病情严重度下的单波段SIF强度,利用对作物冠层几何结构敏感的归一化植被指数(normalized difference vegetation index,NDVI)和重归一化植被指数(re-normalized vegetation index,RDVI)对SIF和PRI进行处理,再利用处理后的SIF和PRI数据构建SISP指数,通过建立传统的光谱指数和SIF、PRI及其组合对小麦条锈病的遥感监测模型,以病情指数(disease index,DI)实测值与预测值之间的决定系数(R2)、均方根误差(RMSE)和相对分析误差(RPD)评价模型精度,进而与SISP指数建立的模型进行比较,分析SISP指数对作物病害遥感监测的有效性。【结果】(1)综合利用SIF和PRI数据能够提高对小麦条锈病的遥感探测精度,3组验证样本数据集中,以PRI和SIF的简单组合PRI+SIF为自变量构建的小麦条锈病监测模型,预测DI值与实测DI值间的R2比单一PRI和SIF至少提高14.0%和1.7%,RMSE至少降低7.1%和3.7%。(2)利用反射率光谱指数NDVI和RDVI处理后的SIF和PRI构建的SISP指数,对小麦条锈病DI的预测精度优于直接利用PRI和SIF组合的各种指数,验证样本数据集中预测DI值与实测DI值间的R2至少提高3.7%,RMSE至少降低9%。(3)以SISP和反射率光谱指数为自变量构建的小麦条锈病多元线性回归(multiple linear regression,MLR)和径向基神经网络(radial basis function neural network,RBFN)模型的精度,高于仅利用反射率光谱指数构建的模型精度,其预测DI值与实测DI值间的R2分别较反射率光谱指数提高13.42%和5.72%,RMSE分别减少29.93%和19.24%,RPD分别提高44.53%和29.80%。【结论】利用NDVI和RDVI处理后的SIF和PRI构建SISP指数,能够减弱作物群体生物量对冠层SIF和PRI信号的影响,提高小麦条锈病的遥感监测精度。
关键词:  日光诱导叶绿素荧光  光化学反射率指数  反射率光谱  小麦条锈病  病害监测  遥感技术
DOI:
分类号:
基金项目:国家自然科学基金青年基金项目(41601467,52079103);国家重点研发计划项目(2017YFE0122400,2016YFB0501501)
Establishment of spectral index based on canopy SIF and PRI and its application in monitoring wheat stripe rust
YAN Jumei, JING Xia, ZHANG Teng,et al
Abstract:
【Objective】Taking advantages of reflectance spectrum in crop biophysics and solar induced chlorophyll fluorescence (SIF) and photochemical reflectance index (PRI) in photosynthetic physiology,the spectral index (SISP) of synergistic canopy SIF and PRI was constructed to improve the detection accuracy of crop diseases using remote sensing.【Method】The single-band SIF intensity under different disease severity of wheat stripe rust was estimated based on three-bands fraunhofer line discrimination (3FLD) algorithm,and SIF and PRI were processed by normalized difference vegetation index (NDVI) and re normalized vegetation index (RDVI),which were sensitive to geometric structure of crop canopy.The SISP index was then constructed using SIF and PRI,and the remote sensing monitoring model of wheat stripe rust was established based on traditional spectral index,SIF,PRI and their combination.Model accuracy was evaluatedby coefficient of determination (R2),root mean square error (RMSE) and relative prediction deviation (RPD) between measured and predicted disease index (DI).The results were also compared with SISP model to evaluate and analyze the effectiveness of SISP index in remote sensing monitoring of crop diseases.【Result】(1) The comprehensive use of SIF and PRI improved remote sensing detection accuracy of wheat stripe rust.In the three validation sample sets,the model using simple combination of PRI and SIF had better prediction in DI with 14.0% and 1.7% higher R2 than PRI and SIF as well as 7.1% and 3.7% lower RMSE.(2) The SISP index constructed by SIF and PRI processed by reflectance spectral indices of NDVI and RDVI had better prediction accuracy for disease index than various indices directly using the combination of PRI and SIF.The R2 between predicted and measured DI was increased by at least 3.7%,while RMSE was decreased by at least 9%.(3) The accuracy of MLR and RBFN models based on SISP and reflectance spectral index was higher than that based on reflectance spectral index only.Average R2 between predicted and measured DI was increased by 13.42% and 5.72%,average RMSE was decreased by 29.93% and 19.24%,and average RPD was increased by 44.53% and 29.80%,respectively.【Conclusion】The SISP index constructed using SIF and PRI processed by NDVI and RDVI reduced influence of crop population biomass on canopy SIF and PRI signals and improved monitoring accuracy of wheat stripe rust based on remote sensing.
Key words:  solar-induced chlorophyll fluorescence  photochemical reflectance index  reflectance spectrum  wheat stripe rust  disease monitoring  remote sensing technology