Reference data

Reference data for global crop mapping

To generate accurate cropland and crop type maps, high quality in-situ reference data is indispensable for both training classification algorithms and validation of the final products. To this end, the WorldCereal Reference Data Module has been set up.

In this module, different datasets from different providers are hosted:

  • Along with standardized metadata, clearly documenting the source and all operations having been applied to the data
  • In a harmonized way, i.e. containing standardized attributes and mapped to a standardized crop type legend

 The Reference Data Module allows users to:

  • View, query and download existing harmonized in-situ reference data
  • Contribute new data to the repository. In order to facilitate this procedure for the user, automated data quality checks are implemented and assistance is foreseen to make the dataset compliant in terms of metadata and properties.

The Reference Data Module is linked to the WorldCereal classification system through means of a STAC catalogue, which can be queried automatically during classification model training.

More information regarding the availability of in-situ reference data at the end of WorldCereal Phase I and the procedures for data harmonization and standardization can be found in this publication. The reference data itself is hosted separately on Zenodo.