Integrating Remote Sensing and Soil Features for Enhanced Machine Learning-Based Corn Yield Prediction in the...

Area of Science:

• Agricultural Science
• Remote Sensing
• Machine Learning

Background:

• Corn yield prediction is crucial for agricultural management and varietal selection.
• Previous studies often lack real-world applicability, focusing on smaller or controlled areas.
• This research addresses yield prediction in a production-scale, rain-fed environment.

Purpose of the Study:

• To identify optimal vegetation indices and abiotic factors for corn yield prediction.
• To determine the most effective corn growth stage for yield estimation using machine learning.
• To evaluate the performance of different machine learning models for corn yield prediction.

Main Methods:

• Utilized high-resolution (6 cm) aerial multispectral imagery.
• Derived 62 predictors including soil properties, slope, spectral bands, and vegetation indices (e.g., GNDRE, NDRE, TGI) across seven corn growth stages (V4-V14/VT).
• Evaluated four machine learning algorithms: linear regression, random forest, extreme gradient boosting, and gradient boosting regressor.

Main Results:

• The random forest model at the V14/VT growth stage achieved the highest accuracy (RMSE of 0.52 Mg/ha).
• Yield estimation at the V6 stage was also found to be feasible.
• Integration of abiotic factors (slope, soil properties) and specific vegetation indices (TGI, HUE, GNDRE) significantly improved prediction accuracy.

Conclusions:

• Machine learning models, particularly random forest, combined with abiotic factors and vegetation indices, can accurately predict corn yield at a production scale.
• Early-season yield estimation is feasible, aiding farmers and crop consultants in planning and decision-making.
• The findings support enhanced farm profitability and sustainability through improved yield prediction.