📄 New paper out in Remote Sensing (MDPI)!

Study area: the Thumb (TB), Southwest (SW), and Southeast (SE) regions of Michigan’s Lower Peninsula, with visited and re-visited fields (Shao et al., 2026 — Figure 1).

We used Sentinel-1 + Sentinel-2 imagery and Random Forest in Google Earth Engine to map cover crop species across three regions in Michigan’s Lower Peninsula.

🔑 Key findings:

  • SAR + optical consistently outperformed single-sensor models
  • Overall accuracy of 60–80% F1 cereal rye: 0.72 any cover crop: 0.77
  • Red-edge indices (CIre) and April–May imagery were the top predictors
  • SAR filled critical cloud gaps during the spring growth window
  • Outperforms the USDA Cropland Data Layer for cover crop detection

These results have important implications for understanding the extent of cover crop adoption across large-scale farming systems.

Work led by Yiwen Shao, with Jennifer Blesh, Haoyu Wang, Preeti Rao, and Meha Jain at the University of Michigan School for Environment and Sustainability.

📖 Open access: Shao et al. (2026), Remote Sensing 18(12), 1933