TY - JOUR
T1 - Prediction of winter wheat nitrogen status using UAV imagery, weather data, and machine learning
AU - Tanaka, Takashi
AU - Gislum, René
PY - 2025/3
Y1 - 2025/3
N2 - The critical nitrogen dilution curve (CNDC) and associated nitrogen nutrition index (NNI) are known to provide valuable information indicating whether the crops are experiencing luxury nitrogen (N) uptake—where they absorb more N than needed for optimal growth— or suffering from N insufficiency, where they fail to meet their optimal growth requirements. The aim of this study was to explore the potential of using UAV-based remote sensing and weather data to quantify NNI in a winter wheat crop. For that purpose, field trials with different N application strategies were conducted over three cropping seasons. The calibrated CNDC used in this study showed a better performance in detecting yield reduction caused by the N insufficiency compared to using a CNDC developed in a previous study (default CNDC). Machine learning models (i.e., random forest and partial least squares regression) were used to predict shoot biomass, N concentration, and NNI. The results showed that machine learning models could predict crop N status at medium or high accuracies (R2: 0.59–0.95). However, the default NNI predictions based on UAV data consistently indicated N insufficiency even when the crop was not suffering from N insufficiency. Whereas the calibrated NNI predictions occasionally could detect a reduction in yield caused by N deficiency. Robustness and scalability of the CNDC have rarely been discussed but based on our findings we suggest testing whether the preferred CNDC should be calibrated for a specific cultivar or region is particularly important when using remote sensing technologies for nondestructive N status measurements.
AB - The critical nitrogen dilution curve (CNDC) and associated nitrogen nutrition index (NNI) are known to provide valuable information indicating whether the crops are experiencing luxury nitrogen (N) uptake—where they absorb more N than needed for optimal growth— or suffering from N insufficiency, where they fail to meet their optimal growth requirements. The aim of this study was to explore the potential of using UAV-based remote sensing and weather data to quantify NNI in a winter wheat crop. For that purpose, field trials with different N application strategies were conducted over three cropping seasons. The calibrated CNDC used in this study showed a better performance in detecting yield reduction caused by the N insufficiency compared to using a CNDC developed in a previous study (default CNDC). Machine learning models (i.e., random forest and partial least squares regression) were used to predict shoot biomass, N concentration, and NNI. The results showed that machine learning models could predict crop N status at medium or high accuracies (R2: 0.59–0.95). However, the default NNI predictions based on UAV data consistently indicated N insufficiency even when the crop was not suffering from N insufficiency. Whereas the calibrated NNI predictions occasionally could detect a reduction in yield caused by N deficiency. Robustness and scalability of the CNDC have rarely been discussed but based on our findings we suggest testing whether the preferred CNDC should be calibrated for a specific cultivar or region is particularly important when using remote sensing technologies for nondestructive N status measurements.
KW - Critical nitrogen dilution curve
KW - Nitrogen nutrition index
KW - Partial least squares regression
KW - Random forest
UR - http://www.scopus.com/inward/record.url?scp=85216800611&partnerID=8YFLogxK
U2 - 10.1016/j.eja.2025.127534
DO - 10.1016/j.eja.2025.127534
M3 - Journal article
SN - 1161-0301
VL - 164
JO - European Journal of Agronomy
JF - European Journal of Agronomy
M1 - 127534
ER -