AAVI: A Novel Approach to Estimating Leaf Nitrogen Concentration in Rice From Unmanned Aerial Vehicle Multispectral Imagery at Early and Middle Growth Stages

Abstract

Accurate and timely monitoring of leaf nitrogen concentration (LNC) in rice is crucial to optimize nitrogen fertilizer management and reduce environmental pollution. Existing vegetation indices (VIs) often perform well for high canopy cover conditions, but their performance becomes poor at early growth stages due to the significant exposure of background materials and the induced spectral mixing effect. This study proposed a novel approach to estimate the LNC at early and middle growth stages of paddy rice by using abundance adjusted VIs (AAVIs) from unmanned aerial vehicle (UAV) multispectral imagery. An AAVI was constructed by combining the traditional VI and the rice abundant from linear spectral mixture analysis of UAV imagery. Subsequently, the performance of AAVIs was evaluated in comparison with traditional VIs derived from all pixels or green pixels for individual growth stages or multiple stages. The results demonstrated that AAVIs exhibited better performance in LNC estimation, regardless of individual stages or across the entire early season. Specially, AACIred-edge showed the best performance among the AAVIs evaluated for LNC estimation. For universal modeling across early stages, the combination of AACIred-edge and AAEVI yielded the highest accuracy (R2 = 0.78, RMSE = 0.26%, and rRMSE = 10.4%) performed remarkably better than the traditional VIs from all pixels or green pixels (R2<0.40). These findings illustrated that the AAVIs have great potential in monitoring nitrogen status at early growth stages with high-resolution aerial or satellite images in the context of precision crop management.

Keywords

• Leaf nitrogen concentration (LNC)
• linear spectral mixture analysis (LSMA)
• unmanned aerial vehicle (UAV)
• vegetation indices (VIs)