Remote sensing approaches for crop nutrition diagnosis and recommendations for nitrogen fertilizers in rice at canopy level

Abstract

ABSTRACT Nitrogen (N) fertilizer management plays a crucial role in high-yield rice production. To choose a well-performing rice N nutrient diagnosis indicator for developing rice production management strategies, this research conducted five field experiments under various N treatments. The results showed that machine learning and stepwise multiple linear regression suggested a strong relationship between vegetation indexes and agronomic indicators (0.70 > R2 > 0.51). A strong correlation was obtained between red-edge based vegetation indexes and agronomic indicators (R2 > 0.40). Additionally, the all-subset regression method results demonstrated that the red-edge basis vegetation indexes were generally applied during different vegetation index combinations. The red-edge basis vegetation indexes reached an approximately 40% contribution in nitrogen nutrient index prediction and an approximately 48% contribution in leaf area index monitoring. Furthermore, this study combined the normalized difference red-edge (NDRE) basis dynamic model to calculate the N dose, which ranged from 106 to 134 kg per hectare in large-scale N management according to the NDRE from Sentinel-2B images, a decrease of approximately 46 kg N ha−1 fertilizer compared with farmers’ practices. Nevertheless, more refinements are needed to ensure that this strategy can be applied to farmers’ yield- and income-enhancing production.