Advancing winter wheat yield anomaly prediction with high-resolution satellite-based gross primary production
Résumé
With global food security increasingly threatened by climate variability, accurate crop yield predictions are essential for sustainable agriculture. Winter wheat, with its long growing season influenced by complex climate and management interactions, presents a significant challenge for current predictive models, which often fail under extreme weather conditions despite their complexity and reliance on extensive interactive predictive variables. Here, we employ Harmonized Landsat Sentinel-2 gross primary production (HLS-GPP) data to predict winter wheat yield anomalies across France. HLS-GPP showed strong spatiotemporal correlation with yield anomalies achieving satisfactory accuracy 1 month before harvest and optimal predictions 2 weeks prior. Including rainfall anomalies further enhanced yield anomaly forecasting under extreme weather conditions, where HLS-GPP better predicted significant losses previously unpredicted by national forecasting systems. Our model offers a more efficient wheat yield forecasting tool by minimizing predictive variables and computational demands, offering a practical tool for adaptive food security planning.