Application of Vegetative Indices for Leaf Nitrogen Estimation in Sugarcane Using Hyperspectral Data

Spectral indices are important tools for monitoring nitrogen levels in plants. This study assessed the potential application of spectral indices in monitoring the nitrogen nutritional status in sugarcane crops. Seven sugarcane varieties and three production environments were studied, with the SP813250 variety cultivated in all three experimental areas. Ammonium nitrate was used as the nitrogen source at doses of 0, 50, 100, and 150 kg ha(-1). Leaf samples for hyperspectral analyses and Leaf Nitrogen Content (LNC) were collected during the maximum vegetative development phase of the crop. Based on reflectance data, 20 spectral indices were calculated and then subjected to simple linear regression (SLR) testing for LNC prediction. In the validation of prediction results, the coefficient of determination (R-2) values, root mean square error (RMSE), and predicted relative error were used as reference. All models were calibrated using the 2012/13 crop data and validated using the 2013/14 crop data. Indices involving the 530-570 nm, 680-750 nm, and 750-1300 nm spectral ranges showed the best performance in model validation. Across all varieties and production environments, the most acceptable indices were: BNi (R-2 > 0.66, RMSE < 3.50 g kg(-)1), GNDVI (R-2 > 0.65, RMSE < 3.67 g kg(-1)), NDRE (R-2 > 0.68, RMSE < 3.18 g kg(-1)), RI-1db (R-2 > 0.69, RMSE < 3.66 g kg(-1)), and VOGa (R-2 > 0.69, RMSE < 3.44 g kg(-1)). The environment significantly influenced the predictive potential for the SP813250 variety, with some cases showing up to a 50% reduction in R-2.

Automated summary

This study assessed the potential application of spectral indices in monitoring the nitrogen nutritional status in sugarcane crops. The results showed that spectral indices involving specific wavelength ranges performed the best in model validation. The BNi, GNDVI, NDRE, RI-1db, and VOGa indices were found to be the most acceptable across all varieties and production environments. The environment significantly influenced the predictive potential for the SP813250 variety.