Quantifying field-level cover cropping in the U.S. Midwest using multi-source satellite data

Cover cropping has been considered a promising solution to reduce soil erosion, enhance nutrient cycling, and suppress weeds to improve agroecosystem resilience. Monitoring cover crop adoption is crucial to enable sustainable agriculture, evaluate outcomes of cover crop programs from federal and private sectors, and design effective management strategies. Traditional field cover crop surveys are labor-extensive and costly, while satellite mapping of cover crops is efficient and cost-effective. Accurate mapping of cover crops using remote sensing is challenging because the remote sensed cover crop signals are affected by multiple factors such as soil background, climatic conditions, and the mixed spectra from the crop residues, cash crops, and other land covers. As satellite imagery's spatial, temporal, and radiometric performance varies from sensor to sensor, it is important to evaluate sensors' capacity to identify cover crop adoption. The impacts of soil properties, climatic conditions, and geographical locations should also be included in the cover crop mapping model for large-scale applications. By comparing the Vegetation Index (VI) time series in fields with and without cover crops, we found that cover cropping results in large VI variations or high VI values in non-growing seasons, which are affected by soil properties, climatic conditions, and geographical locations. For field-level cover cropping, Harmonized Landsat-8 and Sentinel-2 (HLS) significantly outperformed the MODIS because the spatial resolution of MODIS is low, and slightly outperformed PlanteScope because the radiometric calibration of PlanteScope is required and valid PlanetScope data are lacking in winter. The MODIS-Landsat fusion data STAIR (SaTellite dAta IntegRation), which have a similar spatial, temporal, and radiometric performance to HLS, have been adopted to quantify cover cropping across the U.S. Midwest since 2000. Our results provide valuable insights into sensor selection and large-scale mapping of field-level cover cropping, which are useful to stakeholders for regenerative agriculture.