Extract canopy structural variables from multi spectral

Leaf Nitrogen Concentration and Plant Height Prediction for Maize Using UAV-Based Multispectral Imagery and Machine Learning Techniques

To estimate plant height and leaf nitrogen concentration, VIs and spectral bands were trained to random forest model. The model with VIs shows better performance. That is partly attributed to the fact VIs modified original spectral band to weighted one. Interestingly, estimated nitrogen trend did not follow chlorophyll related bands. In my opinion, there are 3 possibilities. The one is that LAI blurs chlorophyll signal on red and red-edge. Some already reported that canopy and leaf spectral are inversely related on PLS coefficient. Also, chlorophyll signal is so narrow that multispectral sensor could not capture precisely. Finally, the relationship between nitrogen and chlorophyll is derived from the fact nitrogen is limiting element on soil. On the experimental setting, some region was fertilized. Thus the relationship would be broken. Again, the saturating threshold implies that nitrogen, as non-limiting element, disturbed retrieval of leaf nitrogen.

Reference

Osco LP, Junior JM, Ramos APM, Furuya DEG, Santana DC, Teodoro LPR, Gonçalves WN, Baio FHR, Pistori H, Junior CAS, Teodoro PE. (2020). Leaf Nitrogen Concentration and Plant Height Prediction for Maize Using UAV-Based Multispectral Imagery and Machine Learning Techniques. Remote Sensing, 12(19), 3237.