【作者】 张金恒；

【导师】 王人潮； 王珂；

【作者基本信息】 浙江大学 ， 土壤学， 2004， 博士

【摘要】 氮素是水稻营养三要素中的主要营养元素，它在氨基酸、蛋白质、核酸、叶绿素和酶等物质的生物合成中起到重要的作用。水稻氮素营养水平不同将在水稻生物物理和生物化学上引起相应的变化。水稻缺氮时叶片数量减小、叶片发育不充分、叶肉细胞小、叶片老化、叶绿素含量降低、叶片由下向上逐渐变黄、含水量和蛋白质含量降低、淀粉和糖等含量上升，氮素营养过剩不仅造成水稻贪青徒长、产量下降，还会造成肥料浪费和环境污染。因此如何有效合理地进行氮素管理，提高水稻氮素利用率十分有意义。氮素管理最终将取决于土壤或作物氮素状况的精确评价，因为作物生长反映了所有氮素源的全氮供应，所以作物氮素状况是适时诊断作物氮素有效性的较好指示器。 传统的水稻氮素营养诊断方法主要是基于植物组织的实验室化学分析。这些分析方法从采集大量的样本、烘干、称重、研磨直到使用有潜在危害性药品进行测试，需耗费大量的时间、人力和物力。传统的野外水稻氮素营养诊断方法主要有肥料窗口法诊断水稻氮素营养和根据叶色诊断水稻氮素营养方法。肥料窗口法不能量化追肥量，还需要常规测试，叶色诊断法不能区分作物失绿是由于缺氮引起的还是由于其它因素引起的。叶色诊断法不仅受到人们对颜色的判断能力差异的影响还受到品种、植被密度，导致土壤氮素状况和叶绿素含量变化的作物胁迫等因素影响。 植物遥感诊断技术涉及到各种植物化学成分的估测，特别是高光谱高分辨率及其巨大数据量的特点使得这一技术在植物的精细分类、农作物的长势监测和农田水肥状况的分析等方面展现了巨大的应用前景。植物生物物理和生物化学特征及参数提取的研究成为高光谱遥感技术最精华和核心的部分。 地面多光谱遥感数据监测大面积水稻氮素状况是有潜力的，并且地面多光谱数据诊断水稻氮素营养的研究又为卫星遥感监测大面积水稻氮素状况提供理论依据。 本文利用高光谱遥感的独特性能，将水稻高光谱反射率及窄波段光谱 中文摘要反射率与水稻氮素营养的实验室化学分析相结合，采用相关分析、回归分析和差异显著性分析方法，研究水稻光谱遥感诊断氮素营养的机理。由于从高光谱窄波段的角度与氮素营养之间建立的关系在不同环境条件、不同作物品种之间具有不稳定性，因此本研究在水稻氮素营养光谱诊断方法的研究上避开了单纯使用高光谱连续波段或者窄波段的方法，从以下方面进行研究:.前人的研究已经证明了反射光谱一阶导数的红边斜率与叶绿素含量相关，日一卜绿素含量又与氮素含量相关;反射光谱一阶导数的红边位置与叶面积指数相关，口一1·面积指数也与氮素含量相关，因此反射光谱一阶导数的两个特征参数:红边斜率和红边位置问接地与氮素含量相关，然而这些相关关系的稳定性受品种、生育期和环境条件的影响。基于这一原理，本文综合考虑水稻上下两功能11一卜片反射光谱一阶导数红边位置和红边斜率，提出了诊断氮索营养新的植被指数，暂命名为“红边肩夹角植被指数”(Rededgeeurve 52飞。uxder Angle vegetation Index)，简称RSAVI。初步证明了该植被指数对不同品种、不同生育期、不同环境条件的水稻氮素营养诊断比较理想。.由于叶片反射光谱不仅受到n一!片化学组分的影响，还受到叶肉细胞结构、水分、叶片表层蜡质等多种物理因素的影响，因此基于矿物质光谱特征连续统去除法的原理及其在植被光谱研究中的一些特点，本文将这种方法引入到水稻鲜n一卜片光谱反射率诊断氮素营养的研究中。研究表明在不同品种、不同生育期以及不同环境条件之间，鲜叶片连续统去除的特征参数一一吸收谷整体面积(A)与氮素营养之间的相关性较为稳定，该特征参数不仅能定性评价氮肥水平，并且还可以定量评价水稻氮素营养。.由于高光一潜窄波段与氮素营养相关性不稳定，因此本文还从叶片宽波段角度出发探索诊断水稻氮素营养的方法，研究表明在不同品种、不同生育期、不同生长环境下存在诊断氮素营养稳定性较好的具有新含意的宽波段组合植被指数。.基于水稻叶片反射光谱研究得出的某些方法在诊断水稻氮素营养的研究中具有可行性，这些研究也为诊断水稻氮素营养新仪器的开发提供了理论依据，但是这些方法只局限于地面研究。是否可以发挥卫星遥感的优势，将其应用于诊断水稻氮素营养的研究，解决这一问题的首先任务是进行地面模拟试验，寻找基于冠层光谱反射率诊断水稻氮素营养的最佳新组合宽 中文摘要波段植被指数。初步证明了TM4/(TM3*TMZ)，TMS/(TMZ*TM3)，(TMI*TM4)/(TMZ*TM3)，(TMI*TMS)/(TMZ*TM3)，(TMI*TM4*TM7)/(TMZ*TM3*TMS)，(TMZ一TM4)/(TMZ+TM4)，(TM3一TM4)/(TM3+TM4)，(TM3一TMS)/(TM3+TMS)用于预测氮素营养比较理想。 概而言之，本次研究提出了“红边肩夹角植被指数”(Red edge curveshoulder Angle vegetation Index简称RsAvl)，初步证明该植被指数在不同品种、不同生育期、不同环境条件下预测氮素营养?更多还原

【Abstract】 Nitrogen is one of the most important nutrients. It is the necessary matter for composing amino acid, protein, nuclear acid, chlorophyll and enzyme. Different nitrogen levels can induce the change of rice physiology. If rice lack nitrogen nutrition, it can not grow adequately, otherwise fertilizer will be wasted and environment will be polluted. So it is very important to manage nitrogen fertilizer effectively. Nitrogen management depends on evaluating nitrogen status of soil and crop accurately. All available nitrogen resource can be showed by crop growth status. So the nitrogen status of a crop is the best indicator for diagnosis nitrogen nutrition.The mostly traditional methods for diagnosis nitrogen nutrition of rice is chemical analysis based on vegetation tissue. Those methods have many disadvantages, for example, sampling, drying, weighting, wasting time and resource, harmful chemical, and so on. In order to research harmless and convenient methods for diagnosis nitrogen nutrition of rice, some methods outdoor were found. Up to now, two harmless and convenient methods were used widely, fertilizer windows methods and foliage color methods. Though they were harmless and convenient, the first one could not estimate how much fertilizer needed quantitatively, and the second one could not tell the true reason of yellowish.Remote sensing has been used in predicting chemical components of vegetation. Especially the higher resolution and magnitude information of hyperspectral data make it possible to identification and classification of vegetation, monitor crop growth status and fertility of cropland. Study on some characters of biophysics and biochemistry for vegetation has become one of the most important tasks of hyperspectral remote sensing. Spectral determination has proved an automatic, quick and nondestructive method for assessment ofnutrition, and the spectral discriminability of the treatments demonstrated the feasibility to evaluate rice N status in field by canopy spectral determination and the potential for detecting rice N status in a wide area by using multispectral remote sensing.In this study, in order to research the mechanism of estimate rice N status, we studied the relationship between hyperspectral data and rice nitrogen concentration. The significant differences for rice nitrogen concentration , hyperspectral reflectance, and multispectral data were studied as well. Results showed that there was instability relationship between hyperspectral data and nitrogen concentration for different varieties, growth stage and environment. So new ideal but not hyperspectral data itself was used to find available methods evaluating rice N status as follows: There are two character parameters for the first derivative of red edge spectral reflectance, red edge slope, and red edge position. The relationship between red edge slope and chlorophyll concentration has been proved by many researchers. And relationship between chlorophyll concentration and nitrogen concentration has been proved too. Relationship has been found between red edge position and leaf area indices, and red edge position was related to nitrogen concentration indirectly. However, those relationships were changing with different varieties, growth stage and environment. It is necessary to find new methods to estimate rice N status using red edge slope and red edge position of the most upper fully expended leaf and the third one. Results indicted a new vegetation index which calculated from red edge slope and red edge position of the first and the third fully expended leaves. And ideal results were drawn using the new vegetation index evaluating rice N status at different varieties, growth phases, and environment. Many factors affect leaf spectral reflectance including foliage chemistry composing, cell structure, water content, and so on. The theory of continuum- removed, which has been used in mineral spectral reflectance, was introduced. And some studies on continuum- removed method s更多还原