A global reference database of crowdsourced cropland data collected using the Geo-Wiki platform.

Area of Science:

• * Earth Observation
• * Geospatial Science
• * Crowdsourcing

Background:

• * Global cropland mapping is essential for food security and environmental monitoring.
• * Existing cropland datasets often lack sufficient validation or global coverage.
• * Remote sensing technologies offer potential for large-scale land cover assessment.

Purpose of the Study:

• * To create a global reference dataset for cropland identification using crowdsourcing.
• * To provide quality-assessed data for validating and comparing remote sensing-based cropland maps.
• * To support the development and training of land cover classification algorithms.

Main Methods:

• * A three-week crowdsourcing campaign using the Geo-Wiki tool.
• * Over 80 international participants reviewed nearly 36,000 sample units for cropland identification.
• * Quality assessment involved a control set of 1,793 student-validated locations and 60 expert validations.

Main Results:

• * A comprehensive global reference dataset on cropland extent was generated.
• * Data includes classifications and average cropland per location/user, split into two parts.
• * Quality assessment data provides insights into crowdsourced data reliability.

Conclusions:

• * The crowdsourced dataset offers a valuable resource for validating and comparing medium- to high-resolution cropland maps.
• * The data can be utilized to train and improve land cover classification algorithms.
• * This approach demonstrates the potential of crowdsourcing for generating reliable geospatial reference data.