Evaluating how lodging affects maize yield estimation based on UAV observations.

Area of Science:

• Agricultural Science
• Remote Sensing Technology
• Agronomy

Background:

• Accurate pre-harvest maize yield estimation is critical for agricultural management.
• Existing remote sensing methods for maize yield lack testing under lodging conditions, leaving their feasibility and impact unclear.

Purpose of the Study:

• To develop a lodging index to quantify lodging severity in maize.
• To assess the impact of lodging on maize yield estimation using remote sensing.
• To improve the accuracy and robustness of maize yield prediction models under lodging stress.

Main Methods:

• Developed a lodging index using RGB and multispectral images from unmanned aerial vehicles.
• Employed a random forest regression (RFR) model incorporating the lodging index.
• Evaluated model performance at key maize growth stages: tasseling, milking, and denting.

Main Results:

• The lodging index effectively quantifies the degree of lodging in maize plots.
• Yield-estimation models incorporating the lodging index showed slightly improved accuracy compared to models without it.
• The RFR model with the lodging index at the denting (R5) stage achieved the best performance (R² = 0.859, RMSE = 1086.412 kg/ha).

Conclusions:

• The developed lodging index is a valuable predictor for maize yield estimation under lodging.
• Incorporating lodging stress variables into remote sensing models enhances the accuracy and robustness of crop yield estimates.
• This approach provides valuable insights for precise crop yield estimation in challenging agricultural conditions.