Maize is essential for food security and income in Colombia, but its production faces challenges such as drought, waterlogging, heat stress, and inadequate agronomic practices. To improve production in the face of climate variability, it is crucial to optimize agronomic practices. This study analyzes maize yield in response to agronomy and climate using machine learning algorithms. The approach employed addresses the explainability of machine learning algorithms, including data extraction, transformation, and loading (ETL), algorithm selection and tuning, and techniques to deepen interpretability. The approach seeks to explain the effects of independent predictor variables and their interactions on maize yield. The case study in Colombia uses 5 years of farm-level data from Colombia’s key maize producing regions. The dataset includes data on yield, agronomic management, terrain, and climate. The results provide findings and recommendations based on the models and data for the department of Córdoba. The use of explainability techniques makes machine learning models in agronomy more transparent, thus improving trust and applicability of data-driven recommendations.
Year of publication | |
Authors | |
Geographic coverage | Colombia |
Originally published | 17 Jan 2025 |
Related organisation(s) | CGIAR - Consortium of International Agricultural Research Centers |
Knowledge service | Metadata | Global Food and Nutrition Security | Climate extremes and food securityResearch and Innovation | Climate-smart agriculture |
Digital Europa Thesaurus (DET) | cerealsDataModellingartificial intelligenceadaptation to climate changemachine learning |