In-season prediction of maize stem lodging risk using an active canopy sensor

Lodging is a major problem in maize (Zea mays L.) production worldwide. An analytical lodging model has previously been established. However, some of the model inputs are time consuming to obtain and require destructive plant sampling. Efficient prediction of lodging risk early in the season would be beneficial for management decision-making to reduce lodging risks and ensure high yield potential. Remote sensing technology provides an alternative method for fast and nondestructive measurements with the potential for efficient prediction of lodging risks. The objective of this study was to explore the potential of using an active canopy sensor for the early prediction of maize stem lodging risk using simple regression and multiple linear regression (MLR) models. The results indicated that the MLR models using active canopy sensor data together with weather and management factors performed better than simple regression models using only sensor data for predicting maize stem lodging indicators. Similar results were achieved either using regression models to predict the maize stem lodging risk indicators directly or using the regression models to predict lodging related plant parameters as inputs to a process-based lodging model to predict lodging risk indicators indirectly, although the latter approach using MLR models performed slightly better. A medium planting density (7.0 plants m^-2) and 240 kg ha^-1 N rate would be suitable in the study region, and the recommendations may be adjusted according to different weather conditions. It is concluded that maize stem lodging risks can be predicted using active canopy sensor data together with weather and management information at V8 stage, which can be used to guide in-season management decisions. Additional research is needed to evaluate the potential of using unmanned aerial vehicles and satellite remote sensing technologies in conjunction with machine learning methods to improve the prediction of lodging risks for large scale applications.