Rapid Crop Cover Mapping for the Conterminous United States

Area of Science:

• Agricultural Science
• Remote Sensing
• Environmental Monitoring

Background:

• Accurate and timely crop cover maps are crucial for environmental planning and research.
• Existing methods for generating crop maps are often retrospective, limiting their utility for near-real-time applications.

Purpose of the Study:

• To adapt a historical crop classification model (CCM) for rapid, in-season generation of crop cover maps.
• To assess the feasibility of producing conterminous United States-wide crop maps by September at a 250m resolution.

Main Methods:

• Modified a previously developed CCM by incrementally removing variables to enable rapid mapping during the growing season.
• Implemented a 'two-model mapping' approach, separating crop type classification from 'other' crop classification.
• Utilized a binary model for 'other' crops and a specific crop type model for nine distinct crops.

Main Results:

• Successfully generated conterminous United States-wide rapid crop cover maps at 250m resolution by September.
• Achieved low training errors (0.8% and 1.5%) and acceptable test errors (5.5% and 6.4%) for the two models.
• Demonstrated spatial mapping accuracies for annual maps exceeding 70%.

Conclusions:

• The modified CCM approach is viable for rapid crop cover map generation during the growing season.
• The 'two-model mapping' strategy effectively classifies specific crops and distinguishes 'other' crops.
• This methodology holds significant potential for providing timely, high-resolution crop cover data for environmental applications.