EARLY-SEASON CROP CLASSIFICATION WITH PLANET FUSION

Accurate early-season classification of crops plays a critical role in agricultural monitoring, enabling resource allocation decisions to be made earlier and with greater confidence. Our study introduces a novel early-season crop classification method enhanced with geographical and climate zone information, utilizing gap-free, cloud-free, harmonized Planet Fusion time series data. In a country with extensive climatic diversity like France, our experiments, spanning two different years, demonstrated that a variety of crop types-including winter cereals, maize, and oil-seed rape-could be accurately classified with an F1 score of approximately 80 percent, at least 50 days before the estimated average harvest date. This study illustrates the effectiveness of novel deep learning approaches for integrating diverse data sources in remote sensing, highlighting the potential for significant advancements in early season crop type identification and predictive agricultural management.