Estimating total leaf nitrogen concentration in winter wheat by canopy hyperspectral data and nitrogen vertical distribution

Abstract

The use of remote sensing to monitor nitrogen (N) in crops is important for obtaining both economic benefit and ecological value because it helps to improve the efficiency of fertilization and reduces the ecological and environmental burden. In this study, we model the total leaf N concentration (TLNC) in winter wheat constructed from hyperspectral data by considering the vertical N distribution (VND). The field hyperspectral data of winter wheat acquired during the 2013–2014 growing season were used to construct and validate the model. The results show that: (1) the vertical distribution law of LNC was distinct, presenting a quadratic polynomial tendency from the top layer to the bottom layer. (2) The effective layer for remote sensing detection varied at different growth stages. The entire canopy, the three upper layers, the three upper layers, and the top layer are the effective layers at the jointing stage, flag leaf stage, flowering stages, and filling stage, respectively. (3) The TLNC model considering the VND has high predicting accuracy and stability. For models based on the greenness index (GI), mND705 (modified normalized difference 705), and normalized difference vegetation index (NDVI), the values for the determining coefficient (R2), and normalized root mean square error (nRMSE) are 0.61 and 8.84%, 0.59 and 8.89%, and 0.53 and 9.37%, respectively. Therefore, the LNC model with VND provides an accurate and non-destructive method to monitor N levels in the field.