RapidCrops: A pan-European label dataset for large-scale crop classification

Under the AI4SoilHealth project we have created a dataset (“RapidCrops”) to support the automatic mapping of crop types across Europe. Crop type information is essential for monitoring soil health as it provides  systematic insights into crop rotations over time and supports efforts to detect other cropping practices that affect soil health (e.g. tillage & cover crops). The RapidCrops dataset provides approximately 99M agricultural parcel boundaries with harmonised crop type information across a wide spatio-temporal extent; with coverage across seven EU countries for 5-7 years. Based on parcel boundaries and crop type information reported under the EU IACS programme, our methodology seeks to improve the usability of the parcel boundaries without diluting their integrity. Additional attributes are provided to support the use of the data in ML workflows; especially for those leveraging EO data. The dataset builds on top of the EuroCrops [1,2] crop type harmonisation initiative and the fiboa [3] open data standard for parcel boundaries. For enhanced access to the data, the dataset is also made freely available on Source Cooperative [4]. The dataset was also utilised under the Horizon Europe project Open-Earth-Monitor to perform pan-European crop identification across 51M parcels from the year 2022; classifying each parcel into one of 29 crop types [5]. Please see the license terms for underlying datasets below: Data source Data licensing terms Austria INSPIRE public access license & CC-BY-AT 4.0 Denmark INSPIRE public access license & INSPIRE no conditions & CC0 1.0 Universal France Custom open license Germany Custom open licenses: NRW, Brandenberg, LS Netherlands INSPIRE public access license & INSPIRE no conditions & Dutch creative commons license Portugal CC BY 4.0 Spain Custom open license