Prediction of N, P and K contents in sugarcane leaves by VIS-NIR spectroscopy and NPK interaction effect

The content of N, P and K in sugarcane leaves at seedling stage, tillering stage and elongation stage were detected rapidly by visible - near infrared spectrophotometer. A total of 123 leaf sample s' spectra were collected during the three growth periods. After the outliers were detected, 117 valid samples were obtained and spectral data of all samples were preprocessed. Using the spectral data processed by CARS-PCA as an independent variable, a 6-fold cross-validated PLS model for N, P, and K content was established. The R² of the CARS-PCA-PLS model for N, P, K prediction were 0.8591, 0.6769, and 0.9321, respectively. Correlation analysis of the predicted N, P, and K contents were further implemented to explore the interaction effect between NPK. To simulate the interaction effect between those three major nutrients, 19 factors include possible linear, quadratic, cubic relations between NPK were assumed, multi-factor cubic polynomial regression PLS and MLR correction models were established from those factors. In the modified MLR model, the determinants of N, P and K were respectively 0.8908, 08019 and 0.9139, which improved the performance of the model by 3.7%, 18.5% and 1.3% compared with the CARS-PCA-PLS model which only based on the spectral reflectance data. The results show that the application of visible-near-infrared spectrum combined with interaction effects can effectively predict the content of N, P and K in the sugarcane growing stage. It provides important guiding significance for rapid real-time monitoring of sugarcane growth and fertility management.