Estimating Winter Wheat Nitrogen Vertical Distribution With Bidirectional Canopy Reflected Spectrum

Crop nitrogen is a critical variable for crop yield and quality. According to the nitrogen content of crop, the rational use of nitrogenous fertilizer could give rise to the utilization ratio of fertilizer and reduce environmental impact. The vertical distribution of crop nitrogen is increased with plant height, the bottom layer of leaves will turn yellow when under condition of nitrogen deficient. Traditional field nutrient detection and remote sensing inversion mainly focus on crop surface layer, the lack of a method to detect the middle and bottom layer’s nitrogen need to be resolved. In this study, foliage nitrogen vertical-layer distribution inversion methods were studied based on the data of winter wheat ground hyperspectral data and the vertical distribution of foliage nitrogen.In this paper, the main content and results are as follow:(1) This study proposed a method which could estimate winter wheat nitrogen vertical distribution based on continuous wavelet transform (CWT) and bidirectional canopy reflected spectrum. We estimated vertical distribution of winter wheat nitrogen with support vector machine (SVM), partial least square (PLSR), and least-square support vector machine (LS-SVM) models, respectively, using continuous wavelet coefficients of multi-angle hyperspectral reflectance and corresponding foliage nitrogen density (FND). For each method, three models with FND as the dependent variable and continuous wavelet coefficients of hyperspectral reflectance at corresponding angles as the explicative variables were established. There are on signification differences between three different layers in modeling experiment. Independent model verification showed that PLSR models performed more stable than other models at each layer, especially at middle layer and bottom layer. It can be concluded that the pretreatment method of CWT is better in selecting information which is sensitive to nitrogen. The comparative result of FND regression methods indicates that PLSR algorithm is more suitable for FND estimation than others. The method which combines CWT and PLSR is capable of estimating winter wheat nitrogen vertical distribution.(2) The objective of this study was to discuss the method of estimating winter wheat nitrogen vertical distribution by exploring bidirectional reflectance distribution function (BRDF) data using partial least square (PLSR) algorithm. The canopy reflectance at nadir, ±50°and ±60°; at nadir, ±30°and ±40°; and at nadir, ±20°and ±30°were selected to estimate foliage nitrogen density (FND) at upper layer, middle layer and bottom layer, respectively. Three PLSR analysis models with FND as the dependent variable and vegetation indices at corresponding angles as the explicative variables were established. The impact of soil reflectance and the canopy non-photosynthetic materials was minimized by seven kinds of modifying vegetation indices with the ratio R700/R670.The estimated accuracy is significant raised at upper layer, middle layer and bottom layer in modeling experiment. Independent model verification selected the best three vegetation indices for further research. The research result showed that the modified Green normalized difference vegetation index (GNDVI) shows better performance than other vegetation indices at each layer, which means modified GNDVI could be used in estimating winter wheat nitrogen vertical distribution.The result of this study provide theoretical basis for diagnosis of crop nitrogen vertical distribution and early and timely detection of nutrient status, and also provide the reference for exploring bidirectional reflectance distribution function (BRDF) data.