|SHRESTHA, AMRIT - Mississippi State University|
|BHEEMANAHALLI, RAJU - Mississippi State University|
|SAMIAPPAN, SATHISH - Mississippi State University|
|CZARNECKI, JOBY - Mississippi State University|
|MCCRAINE, CARY - Mississippi State University|
|REDDY, RAJA - Mississippi State University|
|MOORHEAD, ROBERT - Mississippi State University|
Submitted to: Frontiers in Plant Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/26/2023
Publication Date: N/A
Interpretive Summary: Corn is one of the important cereal crops, which is cultivated across the globe and its yield enhances with increasing fertilizer application rate. However, over-fertilization to improve yield has resulted in negative consequences like water contamination and deterioration of soil health. To achieve the maximum yield, the fertilizer requirement needs to be applied at the right rate and at the right time based on the crop's seasonal needs, and variable rate technologies (VRT) have that potential. Variable rate technology requires a prescription map for input decisions which are generally based on management zone created by intensive soil sampling which is an expensive procedure and is not feasible throughout the crop-growing season. Therefore, in-season estimation of yield can improve management decisions for agricultural inputs and can be estimated by monitoring canopy optical properties. Different vegetation indices (VIs) have been reported to be highly correlated with crop yield at the different phenological stages. However, none of the studies, to our best knowledge, has reported suitable VIs that can be used throughout the season to predict yield in corn. A field experiment was conducted for three consecutive years 2019-2022 in the R. R. Foil Plant Science Research Center, Mississippi State University to i) identify the potential VIs that have a significant correlation with yield across the growing season, ii) identify the suitable phenological stage and iii) reliable VI for yield estimation under rainfed environments. Finally, a machine learning model based on the random forest was applied to identify the best indices to estimate yield at different phenological stages.
Technical Abstract: Precision agriculture, using unmanned aerial systems (UAS), provides high temporal and spatial resolution for crop monitoring and informed management decisions to improve yields. However, traditional in-season yield prediction methodologies are often inconsistent and inaccurate due to variations in soil types and environmental factors. This study aimed to identify the best phenological stage and vegetation index (VI) for estimating corn yield under rainfed conditions. Multispectral images were collected over three years (2020-2022) during corn growth, and over 50 VIs were analyzed. Results showed that 25 VIs showed significant correlations (r = 0.7) with yield over three years. Thirteen VIs showed a significant correlation with yield in one out of three years, nine in two years, and five in all three years. Further, combined correlation and random forest analyses between yield and VIs led to the identification of consistent VIs for corn yield prediction. In a three-year study, 5 VIs showed significant correlations with yield over three years, among them with leaf chlorophyll index (LCI), MERIS terrestrial chlorophyll index (MTCI), and modified normalized difference vegetation index at 705 (mND705) being the most consistent predictors of corn yield when recorded around the reproductive stage (70-80 days after planting). The best correlation with yield was achieved by combining red, rededge, and NIR-based VIs. This study demonstrates the dynamic nature of canopy reflectance and the importance of considering growth stages, environmental conditions, and years for accurate corn yield prediction.