Corn-Soy Data Layer: US Midwest Crop Maps 1999-2018

The Corn-Soy Data Layer (CSDL) provides 30-meter resolution maps of corn and soybean crops across the US Midwest from 1999 to 2018. Researchers from Stanford University created it by training a random forest classifier on Landsat satellite imagery using the Cropland Data Layer (CDL) for labels. The dataset is hosted on Google Earth Engine and Zenodo.

Use Cases

• Analyzing long-term crop rotation patterns based on annual corn-soybean maps from 1999-2018.
• Validating and filling gaps in earlier Cropland Data Layer (CDL) maps for Midwestern states.
• Studying the relationship between satellite-derived vegetation indices and crop outcomes.
• Forecasting agricultural production at a regional scale using historical crop type data.

Strengths

• Covers a 20-year time range (1999-2018) for longitudinal analysis.
• Provides 30-meter spatial resolution for field-level monitoring.
• Includes validation against CDL data and county-level crop statistics.
• Publicly hosted on Google Earth Engine for cloud-based analysis.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Focus is limited to corn, soybean, and an aggregated 'other crops' class in the US Midwest.

Provenance

Source

Stanford University, Sherrie Wang

Collection Method

Random forest classification of Landsat 5, 7, and 8 Surface Reflectance time series, trained on CDL data.

Time Range

1999-2018

Geography

US Midwest