Global crop calendars of maize and wheat in the framework of the WorldCereal project

Crop calendars provide valuable information on the timing of important stages crop development such as planting or Start Season (SOS) and harvesting dates End (EOS). This is critical for many monitoring applications crop-type mapping, condition monitoring, yield estimation forecasting. Spatially detailed provides an asset in this respect, it allows algorithms to account specific local circumstances while also maximizing their robustness global applicability. Existing calendar products, produced by Group Earth Observations’ Global Agricultural Monitoring (GEOGLAM) Monitor, United States Department Agriculture Foreign Service (USDA-FAS), Food Organization (FAO), European Commission Joint Research Center’s Anomaly hot Spots Production (ASAP), generally only at national subnational level. In work, we present gridded SOS EOS maps wheat maize that represent calendars’ spatial variability 0.5° resolution. These are generated framework WorldCereal, which a Space Agency (ESA) funded project whose cropland scale 10 m resolution require information. The proposed built leveraging above noted products (Crop USDA-FAS, FAO, ASAP) datasets combined into baseline map sampled train Random Forest algorithm based climatic geographic data. Their evaluation against test data from set aside validation purposes show good performance with (EOS) R2 0.87 (0.92) root mean square error (RMSE) 27 (26) days wheat, showing lowest errors (RMSE <15 days) North America, Central Europe, South Africa, Australia, all areas production trade. Meanwhile, largest between 40 60 occurred regions America close Amazon Africa Congo Basin. case maize, shows 0.88 (0.79) RMSE 24 (28) best performing < 15 located Northern Hemisphere, worst were Brazil, Saudi Arabia India. Additionally, evaluated using simple Land Surface Phenology (LSP) model Sentinel-2 Landsat 8 Observation over known fields. results 0.75 (0.88) 25 (18) 0.80 35 (24) maize. Therefore, presented advancement existing terms capturing coverage reporting accuracy.