T
T
T
Treesearch

Harnessing data-driven approaches to predicting corn yields across different weather zones

Formally Refereed
Download (PDF 3.85 MB): https://research.fs.usda.gov/download/treesearch/80460.pdf

Abstract

Forecasting corn yields with precision across diverse weather zones is vital to supporting food supply resilience and sustainable agriculture. We introduce a data-integrated modeling framework applied across the Greater Mississippi River Basin- one of the largest corn-producing regions in the United States. Using MODIS-based NDVI time series, historical meteorological data, and spatial crop coverage, we investigate the relationships between vegetation health indicators and corn yields under varying climatic conditions. A new vegetative metric is introduced to enhance correlation with yield outcomes. Both statistical regressions and advanced machine learning models-including support vector regression, XGBoost, and Gaussian processes-are employed. Among these, XGBoost demonstrated the highest predictive accuracy. The results underscore the benefits of combining multisource data with nonlinear models for regionally robust and scalable corn yield estimation.

Citation

Xie, Weiwei; Huang, Yanbo; Meng, Qingmin; Tang, Bo; Ouyang, Ying. 2026. Harnessing data-driven approaches to predicting corn yields across different weather zones. Journal of Applied Remote Sensing, 20(02). https://doi.org/10.1117/1.jrs.20.024501
Citations