Abstract: 【Objective】 Leaf nitrogen status is a premise for management and control of precise-using nitrogen in wheat production. Non-destructive and real-time assessment of leaf nitrogen content (LNC) has an important significance for production and management of wheat.【Method】Two field experiments were conducted with three different soil textures (sand, loam and clay), five different nitrogen levels (0, 120, 225, 330 and 435 kg•hm-2) and 3 main wheat cultivars in Henan (Aikang58, Zhoumai22 and Zhengmai366) across growing seasons. High spectral reflectance and LNC of canopy were taken by synchronous measurement during main growth stages of wheat. Compared with the high spectral response differences of canopy LNC in wheat under the three different soil textures, several kinds of hyperspectral indices including difference spectral indices (DSI), ratio spectral indices (RSI) and normalized difference spectral indices (NDSI) with all combinations of two wavebands between 350 and 1 050 nm were calculated, their relationships with LNC were analyzed, and the estimation models were established.【Result】 The experimental results showed that there was an obvious difference in the spectra of canopy reflectance under different nitrogen levels and different growth periods, but the trend was almost consistent. Compared with the spectra of canopy reflectance in the three different soil textures, the performance was clay＞loam＞sand, it could reflect the real-time field growing in wheat. The quantitative relationships between the spectra of canopy reflectance and the associated LNC under the three soil textures were systematicaly analyzed, and the calculated results showed that there was a better correlation between the visible and near-infrared area with the different sensitive band intervals. NDSI (FD710, FD690)，DSI (R515, R460) and RSI (R535, R715) were the best indicators to the integrated modeling of LNC in sand, loam and clay, with the predictive determination coefficient (R2) of 0.88, 0.87 and 0.87. Testing the above better spectral parameters of monitoring models with independent sample in 2010-2011, the results reconfirmed that they were the best indicators, with the predictive determination coefficient (R2) of 0.87, 0.85 and 0.77 respectively, and the root mean square error (RMSE) of 0.31, 0.32 and 0.26, respectively. 【Conclusion】 The monitoring model which used the high spectral parameter of NDSI (FD710, FD690), DSI (R515, R460) and RSI (R535, R715) as independent variables, could be used for better estimation of the LNC of wheat in sand, loam and clay soils.

Key words: wheat , soil texture , leaf nitrogen content , hyperspectral remote sensing , monitoring model