- Timely crop monitoring and yield prediction are essential in guiding management decision making. The aim of the study was to estimate the agronomic traits of paddy rice (Oryza sativa L.) using unmanned aerial vehicle (UAV)-multispectral imaging. A randomized complete block design field experiment with a split–split plot arrangement was set up in the Ruzizi plain, Democratic Republic of Congo (DRC). Spectral imaging data were collected at rice tillering and panicle initiation stages. Predictive analysis of rice agronomic traits was performed using linear and decision tree-based machine learning techniques. Paddy rice trait predictions were critically sensitive to the timing of image acquisition but not largely affected by the model. The most accurate predictions were made at rice panicle initiation stage, with R2 values of 0.62, 0.65, and 0.75 for yield, aboveground biomass, and plant nitrogen (N) uptake, respectively. The visible atmospherically resistant index (VARI), modifiedTimely crop monitoring and yield prediction are essential in guiding management decision making. The aim of the study was to estimate the agronomic traits of paddy rice (Oryza sativa L.) using unmanned aerial vehicle (UAV)-multispectral imaging. A randomized complete block design field experiment with a split–split plot arrangement was set up in the Ruzizi plain, Democratic Republic of Congo (DRC). Spectral imaging data were collected at rice tillering and panicle initiation stages. Predictive analysis of rice agronomic traits was performed using linear and decision tree-based machine learning techniques. Paddy rice trait predictions were critically sensitive to the timing of image acquisition but not largely affected by the model. The most accurate predictions were made at rice panicle initiation stage, with R2 values of 0.62, 0.65, and 0.75 for yield, aboveground biomass, and plant nitrogen (N) uptake, respectively. The visible atmospherically resistant index (VARI), modified chlorophyll absorption in reflective index, and ratio vegetation index, along with near infrared and green bands, played a critical role in predicting paddy rice N uptake and yield. The same spectral features associated with crop height and canopy data were essential for predicting paddy rice aboveground biomass. UAV-multispectral data were able to assess agricultural intensification strategies at field/landscape scale irrespective of soil types, watering regimes, and cultivars. Special consideration should be attributed to VARI, as it enables economical prediction of paddy rice traits. The UAV technologies are therefore reliable tools for monitoring rice production and can be applied in agricultural extension in the DRC.…
