Skip to main content
ARS Home » Southeast Area » Stoneville, Mississippi » Crop Production Systems Research » Research » Publications at this Location » Publication #403706

Research Project: Assessment and Improvement of Soil Health under Modern Cropping Systems in the Mid-Southern United States

Location: Crop Production Systems Research

Title: Multi-stage corn yield prediction using high-resolution (UAV) multispectral data and machine learning models

Author
item KUMAR, CHANDAN - Mississippi State University
item Mubvumba, Partson
item Huang, Yanbo
item DHILLON, JAGMAN - Mississippi State University
item Reddy, Krishna

Submitted to: Agronomy Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/27/2023
Publication Date: 4/28/2023
Citation: Kumar, C., Mubvumba, P., Huang, Y., Dhillon, J., Reddy, K.N. 2023. Multi-stage corn yield prediction using high-resolution (UAV) multispectral data and machine learning models. Agronomy Journal. 13(5):1277. https://doi.org/10.3390/agronomy.
DOI: https://doi.org/10.3390/agronomy

Interpretive Summary: Timely and cost-effective prediction of crop yield is vital for informed decision-making in crop management. Conventional methods of yield estimation are labor-intensive, time-consuming, and costly, and are often conducted at the end of the season. Scientists from USDA-ARS, Crop Production Systems Research Unit, Stoneville, Mississippi; USDA-ARS, Genetics and Sustainable Agriculture Research Unit, Starkville, Mississippi; and Mississippi State University, Mississippi State, Mississippi have evaluated the potential application of Unmanned Aerial Vehicle (UAV) (drone) and Vegetation Indices (VIs) combined with different Machine Learning (ML) models for predicting corn yield during vegetative and reproductive growth stages. Agronomic treatments such as Austrian Winter Peas (AWP) cover crop, biochar, gypsum, and fallow were applied during the non-growing corn season to assess their impact on yield. Several variables derived from multispectral data were used to evaluate their effectiveness in predicting corn yield using various ML models, including Linear Regression (LR), k-Nearest Neighbor (KNN), Random Forest (RF), Support Vector Regression (SVR), and Deep Neural Network (DNN). The red edge spectral band and their VIs were found to be suitable for predicting corn yield at both growth stages. This study confirmed that UAV-derived VIs in conjunction with ML models can produce an adequate corn yield prediction even with a limited number of training samples and could be effectively used in better-informed crop decision-making.

Technical Abstract: Timely and cost-effective crop yield prediction is vital for better-informed crop management decision-making. This study evaluates the efficacy of Unmanned Aerial Vehicle (UAV)-based Vegetation Indices (VIs) coupled with the state-of-the-art Machine Learning (ML) models for corn yield prediction at vegetative (V6) and reproductive (R5) growth stages using a limited number of training samples at the farm-scale. Based on one-year experimental results, different agronomic treatments such as Austrian Winter Peas (AWP) (Pisum sativum L.) cover crop, biochar, gypsum, and fallow during the non-growing corn (Zea mays.) season indicated negligible impact on overall corn yield. Thirty different variables (i.e., four spectral bands: green, red, red edge, and near infrared and twenty-six VIs) were derived from UAV multispectral data collected at V6 and R5 stage to assess their utility in yield prediction. Furthermore, five different ML algorithms including Linear Regression (LR), k-Nearest Neighbor (KNN), Random Forest (RF), Support Vector Regression (SVR), and Artificial Neural Network (DNN) were used for yield prediction. Specifically, Red edge (RE), canopy chlorophyll content index (CCCI), red-edge chlorophyll index (RECI), chlorophyll absorption ratio index (CARI), green normalized difference vegetation index (GNDVI), green spectral band (G), and chlorophyll vegetation index (CVI) were among the most suitable variables in predicting corn yield. The SVR and KNN outperformed other models in corn yield prediction at both growth stages. The corn yield prediction within the fallow field was more accurate than other treatments regardless of the ML model used. The SVR predicted yield for the fallow with a coefficient of determination (R2) and root mean square error (RMSE) of 0.84 and 0.69 Mg/ha at V6 and 0.83 and 1.05 Mg/ha at the R5 stage, respectively. The KNN achieved a higher prediction accuracy for AWP (R2 = 0.69 and RMSE = 1.05 Mg/ha at V6 and 0.64 and 1.13 Mg/ha at R5) and gypsum treatment (R2 = 0.61 and RMSE = 1.49 Mg/ha at V6 and 0.80 and 1.35 Mg/ha at R5). The DNN achieved a higher prediction accuracy for biochar treatment (R2 = 0.71 and RMSE = 1.08 Mg/ha at V6 and 0.74 and 1.27 Mg/ha at R5). For the combined (AWP, biochar, and gypsum) treatment, the SVR produced the most accurate yield prediction with an R2 and RMSE of 0.36 and 1.48 Mg/ha at V6 and 0.41 and 1.43 Mg/ha at the R5 stage. Overall, high ac-curacy was achieved with management specific yield prediction compared to combining the treatments. Furthermore, VIs coupled with ML models could be efficiently used in corn yield prediction at the farm scale even with a limited number of training samples as demonstrated in this study.