Premium
Estimating Wheat Shoot Nitrogen Content at Vegetative Stage from In Situ Hyperspectral Measurements
Author(s) -
Bao Yansong,
Xu Kang,
Min Jinzhong,
Xu Jianjun
Publication year - 2013
Publication title -
crop science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.76
H-Index - 147
eISSN - 1435-0653
pISSN - 0011-183X
DOI - 10.2135/cropsci2013.01.0012
Subject(s) - hyperspectral imaging , mathematics , coefficient of determination , stage (stratigraphy) , correlation coefficient , canopy , mean squared error , remote sensing , botany , horticulture , biology , statistics , geology , paleontology
Timely assessment of crop N content is critical for crop growth diagnosis and precision management to generate higher yield and better quality. The objective of this study was to determine the optimal spectral index and build a retrieval model for diagnosing shoot N content (SNC) of wheat ( Triticum aestivum L.) at vegetative stage using ground‐based hyperspectral reflectance data. Hyperspectral indices were investigated to evaluate their capabilities for wheat N concentration estimation by the Pearson's correlation analysis. The analysis results showed that green normalized difference vegetation index (GNDVI) and the combined spectral index the first derivative of reflectance spectral at 736 nm (D736) × the reflectance at 900 nm (R900)/the reflectance at 720 nm (R720) were most suitable for wheat SNC estimation at vegetative stage. A power model with GNDVI and a linear model with D736 × R900/R720 were appropriate for SNC estimation in vegetative stage. The validation experiments demonstrated that the power model with GNDVI was preferable to the linear mode with D736 × R900/R720 for SNC estimation until the flag leaf stage. However, the linear model with D736 × R900/R720 was better after the flag leaf stage. For wheat SNC assessment at the whole vegetative stage, the linear model with D736 × R900/R720 was the most accurate, of which the root mean square error was 2.391 g m −2 and the correlation coefficient between the measured and estimated SNC was 0.934 ( n = 79).