The Optimum NDSI for Monitoring Canopy Leaf Nitrogen Status in Wheat
会议名称：《亚洲精细农业与计算机农业应用联合大会》
会议日期：2009年
学科分类：082804[工学-农业电气化与自动化] 08[工学] 0828[工学-农业工程] 09[农学] 0903[农学-农业资源与环境] 0901[农学-作物学]
关 键 词：Wheat Canopy LNC LNA the optimum NDSI SAVI Bandwidth
摘 要：Fast and nondestructive monitoring nitrogen status has the great significance on precision nitrogen management in wheat. The primary objective of this study was to explore the optimum wavebands, spectrum indices and quantitative models for estimating leaf nitrogen content （LNC） and leaf nitrogen accumulation （LNA） in wheat by precise analysis on canopy hyperspectral information. On the basis of detailed data from 4-year field experiments under varied nitrogen rates and wheat cultivars, a systematic analysis was undertaken on quantitative relationships of LNC and LNA to NDSI （Normalized difference spectrum index, NDSI） composed of any two wavebands with original reflectance and its derivative within the spectral range of 350-2500 nm, SASI （Soil adjusted spectrum index, SASI） and the selected best spectrum index with different bandwidths. The results showed that the sensitive wavebands of nitrogen status mostly lie in the regions of visible and near-infrared. The LNC monitoring models developed from NDSI （R1350, R700） and NDSI （FD700, FD690） gave decision coefficient（R2） of 0.843 and 0.857, with stand error （SE） as 0.380 and 0.362, respectively. The LNA models based on NDSI （R860, R720） and NDSI （FD736, FD526） had R2 of 0.900 and 0.885, and SE of 1.327 and 1.449. Testing of the LNC models on NDSI （R1350, R700） and NDSI （FD700, FD690） with independent experiment data showed that R2 value is over 0.758 and RRMSE is lower than 0.266. Then the models constructed on SASI （R1350, R700） displayed the best performance with L parameter as 0.09. Furthermore, expanding the bandwidths of NDSI （R1350, R700） at 1nm interval produced LNC models with similar performance within about 20nm bandwidth, over which the statistical parameters of the models became less stable. Validation of the derived LNA equations with the independent dada also found that the models based on NDSI （R860, R720） and NDSI（FD736, FD526） produced estimation accuracy as more over 0.812 with RRMSE less than 0. 253