稻纵卷叶螟为害特征参数成像高光谱反演机理与早期预警研究

Rice leaf folder (Cnaphalocrocis medinalis) is one of migratory insects, causing marked decrease of rice yield. In recent years, it tends to aggravate. Hence, it is necessary to monitor dynamics of rice damages characteristics in the paddy fields. Hyperspectral imaging technology has the advantage of continuous spectra and can provide more detailed information of land surface. However, at present, this kind of information, which adds one dimension in space, is still not deeply understood by scientists, although non-imaging hyperspectral data have been aroused relatively extensive attention. The paddy fields in Agro-ecological Demonstration Garden, belonging to the Nanjing University of Information Science and Technology, are chosen as the experiment sites. In our project, a near-ground hyperspectral imager will be regularly used to obtain multi-temporal hyperspectral images of the leaves and of the canopy of the paddy rice plants. Combined with pest field survey and rice physiological and biochemical parameters, the spectral parameters extracted from obtained images will be employed to retrieve the damage characteristics caused by this kind of pest. Some spectral feature parameters will be extracted, including feature bands sensitive to the pest damage, particular vegetation indices etc. Their spatial variation in the hyperspectral images will also be analyzed, and the causes will be deduced. Retrieval models for the pest damage characteristics, applicable to different rice growth stages, will be built. Temporal characteristics of spectral parameters will be analyzed, using time series of hyperspectral images. According to the dynamical curves of these spectral feature parameters, the outbreak time of rice leaf folder and its time duration will be determined, combined with the correlation information of spectral and retrieved damage parameters. The early-warning signals before the outbreak time will be defined. Our research will improve the comprehension of pest monitoring mechanism by hyperspectral images, provide a new perspective to improve the application of hyperspectral images, and is of great consequence for crop pests dynamical monitoring and forewarning.

稻纵卷叶螟是严重危害水稻生产的迁飞性害虫。及时准确地获取其田间为害情况能够为农田管理提供有效依据。高光谱遥感作为遥感监测的发展趋势，过去的研究集中在非成像高光谱数据，高光谱遥感影像的反演研究尚处于探索阶段。充分挖掘高光遥感谱影像的信息是定量遥感的迫切需求。本研究利用近地成像高光谱仪，获取稻纵卷叶螟胁迫下的水稻叶片和冠层高光谱影像，监测田间虫情和水稻生理生化参数。计算不同胁迫程度下的反射光谱数据，分析它们与虫害为害特征之间的相关性，提取对胁迫敏感的光谱波段，构建适用于各水稻生育期的植被指数和反演模型，实现稻纵卷叶螟为害特征参数的定量反演，分析光谱在空间上的变化特征，并解释原因。研究高光谱特征参数随虫害发生的动态变化规律，提取爆发日期和持续时间，确定早期预警信号。本研究有利于深入理解高光谱遥感影像定量反演虫害的机理，能够为其在宏观尺度广泛应用提供理论依据，对作物虫害监测和预警研究具有重要意义。

充分挖掘近地面高光谱成像遥感信息是农业定量遥感的迫切需求。本项目开展了连续三年的稻纵卷叶螟胁迫试验，设置了五个初始虫量等级：50头、100头、200头、300头、400头/百株。利用近地面成像高光谱仪、便携式地物波谱仪，获得了稻纵卷叶螟胁迫下的水稻叶片和冠层成像和非成像高光谱数据，并对同期水稻的理化参数也进行了测试。研究发现，随着稻纵卷叶螟初始虫量的增加，在绝大多数波段，不同胁迫等级间的水稻叶片和冠层的反射率都呈现显著差异（P<0.05）。尤以近红外波段反射率的差异最为显著，它随着胁迫等级的加大而不断下降。植被指数和高光谱特征参数也随着胁迫程度的不同而产生显著差异。其中植被指数随着胁迫的加剧呈下降趋势。红边位置、蓝边位置和黄边位置发生“蓝移”，尤以红边位置的“蓝移”情况最显著。不仅如此，研究还发现各个胁迫等级中的未被啃食叶片、被啃食叶片中的完好部分和损伤部分也都各自能够体现出胁迫的严重程度。这说明尽管有些叶片或者叶片的某些部分未被啃食，但只要同株水稻的其他叶片受到了胁迫，这些叶片（或部分）的光谱就会产生显著变化，主要表现为近红外波段反射率的大幅下降。这是一个新的研究结果，证明了只要稻田发生了虫害，未被啃食的叶片也已经产生了内部变化，也可以作为虫害遥感监测的目标。这也从另外一个角度证明了受到胁迫后整株水稻的生长发育都受到了影响。研究还发现随着胁迫等级的加大，水稻的理化参数也发生了显著变化，包括株高变矮，卷叶率增加，叶面积、叶绿素含量、净光合速率、产量下降等，并且发现它们与光谱数据具有显著的相关关系。研究筛选出敏感波段、植被指数、光谱参数等，初步构建了数学模型，利用高光谱数据反演水稻理化指标。研究获得了连续三年多个水稻生育期的监测数据，分析了它们的时间变化规律，发现胁迫对水稻发育进程有影响，严重胁迫下的水稻产生早熟现象。该项目的成果将有助于将来在宏观大尺度上得以推广，对快速无损虫害监测有重要意义。