Cover crops have critical significance for agroecosystem sustainability and have long been promoted in the U.S. Midwest. Knowledge of the variations of cover cropping and the impacts of government policies remains very limited. We developed an accurate and cost-effective approach utilizing multi-source satellite fusion data, environmental variables, and machine learning to quantify cover cropping in corn and soybean fields from 2000 to 2021 in the U.S. Midwest. We found that cover crop adoption in most counties has significantly increased in the recent 11 years from 2011 to 2021. The adoption percentage of 2021 is 3.3 times that of 2011, which was highly correlated to the increased funding for federal and state conservation programs. However, the percentage of cover crop adoption is still low (7.2%).  The averaged county-level cover crop adoption rates in 2000-2010 and 2011-2021 are publicly available on Dryad.

We used the STAIR algorithm to fuse Landsat and MODIS to obtain daily 30-m NDVI time series from 2000 to 2021 for all the corn and soybean fields in the Midwest. Cover crop features were extracted by decomposing the STAIR NDVI time series into three components: potential cover crop growth features, soil baselines, and cash crop growth curves. The thresholds for cover crop features were modeled with the inputs of climatic variables, soil properties, and geographic locations. Finally, we compared the cover crop features and thresholds to determine cover crop fields.

We recommend that users consider metrics such as (1) the accuracy of CDL and CSDL and (2) the corn and soybean fraction to determine which states/counties' cover crop maps are of high quality.