What is it about?
The CSM-CERES-Maize model was calibrated in NW Spain to simulate forage maize yield and quality using historical climate data. These simulations trained LightGBM models, achieving high accuracy (R²≈0.85) and highlighting growing season length and radiation as key predictors. The outcome is a user-friendly tool for yield and quality forecasting.
Featured Image
Photo by Ash Willson on Unsplash
Why is it important?
The CSM-CERES-Maize (DSSAT) model was calibrated with field data from Asturias and Galicia and combined with historical weather records to generate a comprehensive dataset of forage maize yield and quality. This dataset integrated both real observations from field trials and synthetic data from DSSAT simulations, which were then used to train a range of machine learning algorithms, with special attention to ensemble models such as LightGBM, Random Forest, and XGBoost. Among them, the LightGBM model achieved the best performance, with high accuracy (R²≈0.85) in predicting dry matter yield, net energy for lactation, and crude protein. The analysis of variable importance revealed that the length of the growing season and cumulative radiation were the most influential predictors. Finally, a web application was developed to make the predictive model accessible to non-specialist users, enabling yield and quality forecasts based on location, cultivar, and sowing and harvest dates.
Perspectives
Writing this article has been a rewarding experience, as developing a predictive model based on hybrid data was a real challenge that required testing several types of machine learning approaches. More importantly, it provided a practical use for almost 25 years of research data that might otherwise have remained in storage without being fully exploited. The culmination of this effort was the creation of a publicly accessible web application, which allows the broader community to benefit directly from the research outcomes.
Silverio Garcia Cortes
University of Oviedo
Read the Original
This page is a summary of: A machine learning approach for estimating forage maize yield and quality in NW Spain, PLOS One, August 2025, PLOS,
DOI: 10.1371/journal.pone.0326364.
You can read the full text:
Resources
Contributors
The following have contributed to this page
