【作者】 周冬琴；

【导师】 曹卫星；

【作者基本信息】 南京农业大学 ， 生态学， 2007， 博士

【摘要】 氮素是植物生长的主要营养元素，在水稻产量品质形成及栽培调控中起到重要的作用。利用遥感技术进行作物长势、氮素营养状况、生产力指标的实时监测与快速诊断是信息农业中的关键技术。本研究以系列田间试验为依托，综合运用现代遥感信息、生长分析、生理生化测试以及数理统计分析等手段，分析了不同条件下水稻冠层多光谱和高光谱特征及其与水稻氮素营养和籽粒品质性状的关系，明确了水稻叶片色素含量、植株氮素状况及品质指标的敏感光谱特征参量及定量反演模型，构建了基于植株特征光谱的水稻氮素营养和籽粒蛋白质含量预测模型，为利用遥感技术监测水稻植株生化组分和籽粒品质等提供了理论依据和技术基础。具体研究内容与结果如下：首先综合运用多光谱和高光谱辐射仪，比较了不同氮素营养水平、不同生育时期和不同品种条件下两种仪器获得的水稻冠层光谱反射率的变化模式。结果表明，水稻冠层多光谱反射率曲线和高光谱在可见光部分几乎重叠，在近红外波段反射率值高光谱略高于多光谱。随着施氮水平的提高，冠层反射光谱在近红外反射平台（750-1300nm）的反射率呈上升趋势，而可见光部分反射率则下降，并且冠层反射光谱的绿峰和红边位置也随着施氮水平的提高分别向蓝光方向（波长变短）和红光方向（长波方向）移动。水稻从分蘖开始，随生育期推进，冠层反射光谱在可见光波段的反射率先降低后升高，以抽穗期反射率最低，随着叶片变黄反射率又增大，且绿光波段的反射峰逐渐消失。而近红外区反射率则表现相反趋势，以开花期为界先上升然后下降，直到成熟前降为最低。不同品种水稻冠层反射光谱随氮肥和生育期的变化趋势相同，只是反射率值的大小不同。系统分析了水稻叶片色素含量与对应冠层多光谱及高光谱反射特征的定量关系。结果表明，不同品种、叶位的叶片色素组分含量均随施氮水平增加而提高。叶绿素a（Chla）、叶绿素b（Chlb）、总叶绿素含量（Chlt）及总叶绿素含量与类胡萝卜素含量之比（Chlt／Car）与可见光波段（350～710 nm）及短波近红外波段（1400～2485 nm）反射率呈负相关，与近红外波段（750～1300 am）反射率呈正相关；类胡萝卜素含量（Car）及叶绿素a与叶绿素b的比值（Chla／Chlb）与冠层单波段反射率的相关趋势与Chla刚好相反。多光谱植被指数RⅥ（660，460）可以较为准确地反演Chla和Chlt；NDVI（610，460）可以反演Chlb；NDⅥ（950，870）可以反演总叶绿素含量与类胡萝卜素含量的比值（Chlt／Car）。高光谱植被指数WI与Chla、Chlb和Chlt均存在极显著线性正相关关系；NDⅥ（610，460）与Chlt／Car存在极显著线性负相关关系。利用不同试验条件下的观测数据对多光谱参数建立的模型进行检验，Ch1a、Ch1b、Ch1t和Ch1t／Car预测值和实测值的拟合斜率（slope）分别为1.062、0.930、1.036和0.983，根均方差（RMSE）均小于20％，表明模型预测值与实测值之间符合度较高，对水稻色素含量具有较好的预测性。综合分析了水稻叶片及地上部植株氮素状况与冠层反射光谱的定量关系，建立了水稻叶片及植株氮素含量和积累量的光谱监测模型。结果显示，不同粳稻品种和生育时期的水稻叶片氮素含量可以用归一化植被指数NDⅥ（1220，710）进行反演，水稻叶片氮积累量可以用比值植被指数RⅥ（1100，560）进行估测，植株氮素积累量的最佳反演光谱参数是差值植被指数DⅥ（1220，870），叶片氮积累量反演的决定系数最高，其次为植株氮积累量和叶片全氮含量。利用不同粳稻品种和水肥处理的观测资料对监测方程进行了检验，根均方差均小于20％，拟合斜率为0.908～1.016，表明预测值与实测值之间符合度较高，对不同栽培条件下的水稻氮素指标具有较好的预测性。在分析水稻叶片糖氮比值随施氮水平变化规律的基础上，提出了不同生育时期水稻叶片糖氮比与冠层高光谱反射特征的定量关系。结果表明，叶片糖氮比与拔节后不同生育时期冠层原始反射率的相关性趋势一致，进一步建立了高光谱参数ND₆₇₂与冠层叶片糖氮比（LCNR）的线性回归方程(LCNR=788.33×ND₆₇₂-4.4326)为水稻冠层叶片碳氮比的最佳监测模型。模型经过不同生育时期数据的交叉测试和独立试验资料的检验，得出对冠层叶片碳氮比的预测精确度范围为0.687-0.986，准确度为0.907-1.126，RMSE为7.07-18.25，表明水稻冠层高光谱特征可以用来定量估测不同栽培条件下叶片碳氮比的变化状况。在分析不同年份、品种和氮素水平下水稻籽粒蛋白质含量和积累量差异的基础上，研究了水稻籽粒蛋白质指标与不同时期冠层反射光谱及叶片和地上部植株氮素状况的关系。成熟籽粒蛋白质含量与不同时期冠层反射光谱的相关分析表明，孕穗期冠层单波段反射率与成熟籽粒蛋白质含量的相关性最高，在16个波段中以760 nm波段反射率与籽粒蛋白质含量的拟合效果最好，建立了水稻成熟籽粒蛋白质含量（GPC）监测模型，GPC=-0.15×DⅥ（1500，950）+3.00．利用不同年份不同品种及不同施氮水平下的观测数据对模型的检验显示，模型预测值与实测值之间符合度较高，对水稻成熟籽粒蛋白质含量具有较好的预测性。此外，利用前期植株氮素状况可以预测成熟期籽粒蛋白质水平，但不同植株氮素指标对籽粒蛋白质的预测能力和最佳时期有所不同。另外，同一植株氮素指标对籽粒蛋白质含量和积累量的监测时期也不完全相同。总体上，不同品种不同氮素水平下水稻从孕穗到灌浆中期的植株氮素状况都能用来反映最终籽粒蛋白质性状。通过分析冠层多光谱水稻籽粒淀粉含量及其他品质性状的相关性，发现利用冠层反射光谱可以定量反演籽粒直链淀粉含量、垩白米率和垩白度，而对其他籽粒品质指标预测并不理想。进一步分析了籽粒品质指标与叶片氮素状况的关系及籽粒品质性状间的相关关系，为水稻籽粒综合品质指标的光谱监测提出了一条间接途径，即在籽粒蛋白质指标光谱监测的基础上，借助籽粒品质指标之间的相关性来预测预报其他籽粒品质性状。更多还原

【Abstract】 Nitrogen is one of the most important nutrients plant growth, and plays key roles in grain yield and quality formation and cultural regulation in rice production. Remote sensing is an important tool for estimating crop growth characters, nitrogen status, yield and quality formation. In this study, a series of field experiments with rice including different varieties, nitrogen and water management practices and population treatments were carried out in five years, canopy spectral reflectance were measured with canopy multi-spectral radiometer （CROPSCAN） and hyper-spectral radiometer （FieldSpec Pro FR2500） over the whole growth periods. The characteristics of canopy multi-spectral and hyper-spectral reflectance under different experimental conditions and their correlations to nitrogen status and grain quality traits in rice were investigated, and the sensitive spectrum parameters and quantitative regression models were established for leaf pigment content, nitrogen status and grain quality traits. This work would provide theoretical basis and key techniques for non-destructive monitoring of nitrogen status and grain quality traits in rice plants with remote sensing technology. The main contents and results of the present study are summarized as follows.The change patterns of canopy reflectance under varied nitrogen rates, different eultivars and different growth stages were investigated based on integrated analysis of canopy multi-spectral and hyper-spectral reflectance features in rice. Results showed that reflectance at near infrared reflected flat increased with increasing nitrogen supply, whereas reflectance at visible band decreased, and green peak and red edge position of canopy reflectance spectra also respectively moved to direction of blue light（short wavelength） and red light （long wavelength）. Reflectance at visible light initially decreased, then increased with growth progress after tillering, and the lowest value appeared at heading. Reflectance increased and reflectance peak also disappeared gradually in course of leaf yellowing. However, reflectance in near infrared range had opposite trend, which initially increased, then decreased to the lowest from anthesis to maturity. The canopy reflectance also differed with eultivars. These results provide theoretical basis for using canopy reflectance spectra to monitor of growth status, nitrogen status and grain qualities in rice. The relationships of leaf pigment contents to canopy spectral reflectance were quantified based on the experiment data. The results showed that the pigment contents of different position leaves increased with increasing nitrogen rates. Leaf chlorophyll a, chlorophyll b, total chlorophyll content and the ratio of total chlorophyll to carotenoid content were negatively correlated to reflectance at 350-710nm and 1400-2485 nm, and positively correlated at 750-1300 nm. Yet carotenoid content and the ratio of chlorophyll a to chlorophyll b content were opposite, negatively correlated to reflectance at 750-1300nm, and positively correlated at 350-710nm and 1400-2485 nm. It is proposed that both leaf chlorophyll a and total chlorophyll can be well monitored by vegetation index RVI（660, 460）, chlorophyll b can be well monitored by NDVI（610, 460）, and the ratio of total chlorophyll to carotenoid content can be well monitored by NDVI（950, 870）. Leaf pigment contents were also significantly correlated with canopy hyper-spectral parameters WI and NDVI（610,460）. The derived equations were established as Chla (mg.g^-1FW) =-0.5718×RVI （660,460） + 2.9335, Chlb (mg.g^-1FW) =-2.0667×NDVI （610,460） + 1.2819, Chlt (mg.g^-1FW)=-0.9075×RVI （660,460） + 4.0158, Chlt/Car =-59.051×NDVI （950,870）+ 4.322. Tests with other independent dataset showed the estimation accuracy of 0.930-1.062, and RMSE of 11.277-19.574% under varied growing conditions. It is concluded that the present model was feasible and reliable for estimating leaf pigment contents in rice with different cultivars and nitrogen levels.The quantitative relationships between leaf and plant nitrogen status and canopy reflectance spectra in rice were investigated. The results showed that NDVI of R₁₂₂₀ and R₇₁₀ is the best parameter for predicting leaf content （LNC）, the RVI of R₁₁₀₀ and R₅₆₀ is the best parameter for predicting leaf N accumulation （LNA）, and the DVI of R₁₂₂₀ and R₈₇₀ is the best parameter for predicting plant N accumulation （PNA）. The three models were tested by the data from independent field experiments. The RMSEs between the estimated and observed values were all below 20%, and slopes were 0.908～1.016, which indicated the monitoring models were feasible and useful for predicting nitrogen content and accumulation in leaves and plant of rice grown under different cultivars and nitrogen levels.Based on the change patterns of leaf sugar and nitrogen content and C/N ratio under different nitrogen supply with growth stages, correlations of leaf C/N to reflectance of single bands, different vegetation indices, derivative indices and parameters normalized by the continuum were analyzed comprehensively. The results showed that there was consistent correlation between C/N ratio and canopy reflectance after jointing in rice. The spectral index of ND₆₇₂ was found to be the best parameter for predicting leaf C/N in rice. The derived equation, LCNRA=788.33×ND₆₇₂-4.4326, was tested with the observed data of seven growth stages in the field experiment. The estimation precision ranged 0.687-0.986, estimation accuracy 0.907-1.126, and RMSE 7.851-18.25, indicating a good fit between the predicted and observed values of leaf C/N. Tests with other independent dataset showed that the estimation precision was 0.857-0.967, estimation accuracy 0.970-1.049, and RMSE 7.07-16.01. Thus, the present hyper-spectral model was feasible and reliable for estimating leaf C/N in rice with different cultivars and nitrogen levels.The relationships of grain protein content to canopy reflectance spectra and leaf and plant N status at different growth stages in rice were quantified based on differences of protein content and accumulation under different years, cultivars and nitrogen rates. The results showed that there were significant negative correlation between grain protein content and canopy spectral reflectance at 460-710 nm and positive correlation at 760-1220 nm after jointing, with best performance from the relationship at 760 nm and booting stage. The differential vegetation index of R₁₅₀₀ and R₉₅₀ was found to be the best parameter for predicting grain protein content （GPC） in rice. The derived equation, GPC =0.15×DVI （1500, 950） + 3, was tested with the observed data from the other independent experiments. There was a good fit between the predicted and observed values of grain protein content. Thus, the present spectral index model is feasible and useful for estimating grain protein content in rice with different cultivars and nitrogen levels. In addition, grain protein index at maturity could be estimated by leaf and nitrogen status at prior stage, but estimating capacity and proper time were different for different nitrogen status. In addition, the proper time for estimating grain protein content and accumulation is different with the same plant nitrogen index. Overall, plant nitrogen status from booting to mid-filling could be used to estimating gratin protein index at maturity for different cultivar and different nitrogen levels.Relationship of grain amylase content and other quality traits to canopy reflectance spectra were investigated using multi-spectral remote sensing data from different experiments. Results showed that amylase content, chalkiness grain percentage and chalkiness degree could be quantitatively estimated by canopy reflectance spectra, but other quality traits were not satisfied. In addition, relationships of grain quality traits to leaf N status and different quality traits were also analyzed for identifying indirect approach to monitor integrated grain quality in rice. The results revealed that indirect estimation of other grain quality straits could be realized based on quantitative monitoring of grain protein content and mutual relationships of grain quality indices.更多还原