基于多种光谱仪的水稻前期植株氮积累量监测

为了明确水稻穗肥施用前地上部植株氮积累量与各光谱仪冠层光谱参数的定量关系，进而为水稻精确追氮提供决策依据。基于不同品种和不同施氮水平的7个水稻田间试验，于分蘖期和拔节期利用ASD Fieldspec FR2500高光谱仪、Cropscan MSR-16多光谱仪和Greenseeker RT100主动光谱仪同时采集冠层光谱反射率，并同步取样测定地上部植株氮积累量，研究基于不同光谱仪构建的植被指数与植株氮积累量之间的关系。结果表明，部分植被指数与水稻地上部植株氮积累量关系密切，基于3种光谱仪构建的水稻地上部植株氮积累量监测模型的稳定性和适用性有较大差异。对于ASD高光谱仪，虽然基于差值植被指数（760，740）能较好估测植株氮积累量，拟合模型决定系数R2为0.79，但模型检验效果较差，其R2和均方根误差分别为0.15和2.11 g/m2；对于Cropscan多光谱仪，差值植被指数（760，710）能较好反演植株氮积累量，拟合模型的R2为0.94，模型检验的R2和均方根误差分别为0.94和0.76 g/m2；Greenseeker主动光谱仪的归一化植被指数（770，660）对地上部植株氮积累量的反演效果最好，拟合模型的R2为0.97，模型检验的R2和均方根误差分别为0.97和0.62 g/m2。研究结果可为水稻前期植株氮积累量监测过程中的光谱仪选择提供参考，为水稻精确追氮管理提供技术支撑。

Monitoring plant nitrogen accumulation with different canopy spectrometers at early growth stages in rice

The objective of this paper is to quantify the relationship between plant nitrogen accumulation (PNA) and canopy spectral reflectance at early growth stages in rice, and provide a technical support for decision making on nitrogen dressing management. Based on seven field experiments with different rice cultivars and nitrogen rates in five different growing seasons, the canopy spectral reflectance during the stages of tillering and elongation were measured with three different canopy spectral sensors (Fieldspec FR 2500, MSR-16 and GreenSeeker RT 100), and the plants were sampled for PNA measurement simultaneously. Then the relationships beween PNA and vegetation indices were analyzed. The results showed that some vegetation indices were closely related to PNA, but there were significant differences among the estimation models based on different canopy sensors. For the hyperspectral sensor (Fieldspec FR 2500), DVI(760, 740) had a good performance on model development, with determination coefficient (R2) of 0.79, but the performance of model testing was poor, with R2 and RMSE of 0.15 and 2.11 g/m2, respectively. For the multispectral sensor (MSR-16), DVI(760, 710) could be used for PNA estimation, with R2 of 0.94 for model development, and with R2 and RMSE of 0.94 and 0.76 g/m2 for model testing, respectively. NDVI(770, 660) from the active spectral sensor (RT 100) gave the best estimation of PNA, with R2 of 0.97 for model development, and with R2 and RMSE of 0.97 and 0.62 g/m2 for model testing, respectively. These results would be helpful for selection of proper spectral sensors in PNA monitoring during early growth stages of rice.