Machine learning algorithms for maize yield prediction with multispectral imagery: Assessing robustness across varied growing environments

Authors: SAPKOTA, BALA RAM - Texas Agrilife Research; BAATH, GURJINDER - Texas Agrilife Research; Flynn, Kyle; Adhikari, Kabindra; Hajda, Chad; Smith, Douglas; MURRAY, SETH - Texas A&M University

Submitted to: Science of Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/4/2025
Publication Date: 8/5/2025
Citation: Sapkota, B., Baath, G., Flynn, K.C., Adhikari, K., Hajda, C.B., Smith, D.R., Murray, S.C. 2025. Machine learning algorithms for maize yield prediction with multispectral imagery: Assessing robustness across varied growing environments. Science of Remote Sensing. https://doi.org/10.1016/j.srs.2025.100267.
DOI: https://doi.org/10.1016/j.srs.2025.100267

Interpretive Summary: Timely and reliable crop yield prediction is essential for optimizing resource allocation and crop management decisions. Multispectral imagery, especially when acquired via Unmanned Aircraft Systems (UAS), can provide an on-demand and cost-effective approach to crop yield estimation. However, traditional methods relying solely on vegetation indices (VIs) for yield prediction face challenges such as saturation in dense canopies, like those of corn, and inconsistencies due to environmental variability. This study explores the potential use of canopy spectral reflectance, directly applied in conjunction with advanced statistical methods such as machine learning (ML), to address these limitations and produce more consistent and reliable yield predictions. The performance of calibrated reflectance-based ML models was evaluated for corn yield predictions across various crop growth phases and compared to VI-based models as well as hybrid models utilizing both reflectance bands and VIs. The research utilizes corn yield data from diverse growing environments, comprising seven corn planting dates tested across three field locations over two years. Our findings indicate that direct incorporation of canopy reflectance bands into ML models, bypassing the complexities associated with VIs, can lead to more consistent, timely, and reliable corn yield predictions across varied growing environments.

Technical Abstract: Timely and reliable crop yield prediction is essential for optimizing resource allocation and crop management decisions. Multispectral imagery, especially when acquired via Unmanned Aircraft Systems (UAS), can provide an on-demand and cost-effective approach to crop yield estimation. However, traditional methods relying solely on vegetation indices (VIs) for yield prediction face challenges such as saturation in dense canopies, like those of corn (Zea Mays L.), and inconsistencies due to environmental variability. This study explores the potential use of canopy spectral reflectance, directly applied in conjunction with advanced statistical methods such as machine learning (ML), to address these limitations and produce more consistent and reliable yield predictions. The performance of calibrated reflectance-based ML models was evaluated for corn yield predictions across various crop growth phases and compared to VI-based models as well as hybrid models utilizing both reflectance bands and VIs. The research utilizes multispectral imagery and corn yield data from diverse growing environments, comprising seven corn planting dates tested across three field locations over two years. Five ML algorithms were implemented for each growth phase dataset using k-fold cross-validation with 70% data, and the model performance was validated externally against an independent 30% data. Among the ML algorithms tested, the Extra Trees (ET) regressor showed superior performance at predicting corn yield for most corn growth phases. Calibrated reflectance-based models consistently outperformed VI-based models at mid-vegetative (R2 = 0.71 vs. 0.50), late vegetative (R2 = 0.85 vs. 0.80), flowering (R2 = 0.83 vs. 0.60), and mid-reproductive (R2 = 0.83 vs. 0.74) growth phases. Hybrid models combining spectral bands achieved the greatest prediction accuracy but required extensive data preprocessing and computational re-sources. Our findings indicate that direct incorporation of canopy reflectance bands into ML models, bypassing the complexities associated with VIs, can lead to more consistent, timely, and reliable corn yield predictions across varied growing environments.