Randomized trial evaluates UAV-derived multispectral imagery to predict physiological health states in tropical papaya, indicating enhanced monitoring capabilities.
Key Points
This research aims to determine the effectiveness of UAV multispectral imagery in predicting physiological health states of tropical papaya.
Field observations on 103 papaya trees in Yogyakarta, Indonesia.
Integrated spectral predictors from UAV imagery with physiological measurements using k-means clustering for intraplant state classification.
Developed and validated a Random Forest model to classify health states based on selected multispectral predictors.
Achieved a training accuracy of 0.958 and testing accuracy of 0.903.
Key spectral predictors identified include GBNDVI_max and MGRVI_std, showing stronger associations with stomatal conductance.
K-means clustering resulted in three physiologically interpretable states: Healthy, Moderate, and Stressed.