Estimating leaf nitrogen accumulation in maize based on canopyhyperspectrum data

摘要

Leaf nitrogen accumulation (LNA) has important influence on the formation of crop yield and grain protein. Monitoring leaf nitrogen accumulation of crop canopy quantitively and real-timely is helpful for mastering crop nutrition status, diagnosing group growth and managing fertilization precisely. The study aimed to develop a universal method to monitor LNA of maize by hyperspectrum data, which could provide mechanism support for mapping LNA of maize at county scale. The correlations between LNA andhyperspectrum reflectivity and its mathematical transformations were analyzed. Then the feature bands and its transformations were screened to develop the optimal model of estimating LNA based on multiple linear regression method. The in-situ samples were used to evaluate the accuracy of the estimating model. Results showed that the estimating model with one differential logarithmic transformation ((teP) of reflectivity could reach highest correlation coefficient (0.889) with lowest RMSE (0.646 gm~(-2)), which was considered as the optimal model for estimating LNA in maize. The determination coefficient (R~2) of testing samples was 0.831, while the RMSE was 1.901 gm"2. It indicated that the one differential logarithmic transformation of hyperspectrum had good response with LNA of maize. Based on this transformation, the optimal estimating model of LNA could reach good accuracy with high stability.