Amazirh A., Chehbouni A., Kreiner A., Adeniyi O.D., Zeleke G.A., Bouras E., Chakir A., Wheeler J., Barnieh B.A., Khechba K., Zoltan S. (2026). Seasonal field-scale wheat yield forecasting using XGBoost with radar, optical, and weather data in Morocco. International Journal of Applied Earth Observation and Geoinformation, 01/05/2026, vol. 149, p. 105242.
https://doi.org/10.1016/j.jag.2026.105242
https://doi.org/10.1016/j.jag.2026.105242
| Titre : | Seasonal field-scale wheat yield forecasting using XGBoost with radar, optical, and weather data in Morocco (2026) |
| Auteurs : | A. Amazirh ; A. Chehbouni ; A. Kreiner ; O.D. Adeniyi ; G.A. Zeleke ; E. Bouras ; A. Chakir ; J. Wheeler ; B.A. Barnieh ; K. Khechba ; S. Zoltan |
| Type de document : | Article |
| Dans : | International Journal of Applied Earth Observation and Geoinformation (vol. 149, May 2026) |
| Article en page(s) : | p. 105242 |
| Langues : | Anglais |
| Langues du résumé : | Anglais |
| Catégories : |
Catégories principales 08 - ALIMENTATION ; 8.3 - Politique et Sécurité AlimentaireThésaurus IAMM SECURITE ALIMENTAIRE ; CEREALICULTURE ; BLE ; RENDEMENT DES CULTURES ; MODELE ; PREVISION DE RENDEMENT ; MAROC |
| Résumé : | Accurately predicting wheat yields at the field scale is critical for ensuring food security and enabling decisionmakers to plan imports and ensure adequate food supplies, particularly in Morocco, where agriculture contributes 11 to 18% of the national gross domestic product. This study explores the potential of integrating Sentinel-1 radar and Senintnel-2 optical with ERA5-Land meteorological data to predict wheat yields at the field level in Morocco. The study implemented the XGBoost model optimized via Bayesian hyperparameter tuning to predict wheat yields across various soil types and agro-climatic zones in the Morocco's wheat-growing regions. The model performed well three months before harvest, achieving a coefficient of determination (R2) of 0.81, a root mean square error (RMSE) of 0.67 t/ha, and a mean error of 0.51 t/ha. Feature importance analysis highlights cumulative precipitation from sowing to heading, the normalized difference vegetation index (NDVI), interferometric coherence, and air temperature as key factors influencing yield predictions. These results demonstrate the effectiveness of combining remote sensing and meteorological data with machine learning for robust, early and accurate wheat yield forecasting-providing crucial decision-making help for national food security and agricultural management. |
| Cote : | En ligne |
| URL / DOI : | https://doi.org/10.1016/j.jag.2026.105242 |


