Satellite Mapping of Tillage and Cover Crop Practices at Field Level for the US Corn Belt

Abstract:

Tillage and cover crop practices directly affect soil health and agroecosystem productivity. Past studies show the potential of reduced/no-tillage and adoption of cover crops in helping improve soil health, reduce soil disturbance, and increase economic benefits. However, continuous observations over a large-scale area combined with modeling work are absent to further confirm these benefits for sustainable agriculture. Only statistical surveys and census data on the adoption of these management practices are available from the USDA, which are not spatially explicit and with low spatiotemporal resolution. Therefore, mapping tillage and cover crop practices at the field level using satellite remote sensing provides an opportunity to fill this data gap. Compared to traditional labor-intensive field surveys, decades-accumulated satellite data has the potential to map these practices with low cost . In this study, we mapped both tillage and cover crop practices at field level in the US Corn Belt based on time-series information of multi-source satellite data. Specifically, we combined spectral-based tillage index, soil background reflectance, and soil moisture information to develop machine learning methods that also lend process-based understanding to map tillage practice at the field level for the whole Corn Belt region. We also used a time series of spectral information and machine learning methods to derive adoption of cover crops at the field scale. We conducted thorough validation of our results using the statistical survey data from the USDA, field-level management data from farmers, and self-collected field-level survey data to demonstrate the performance of our products. The validation results show that our product has a high performance in capturing the spatial pattern and temporal trend of tillage and cover crop practices, and achieves much higher accuracy than the existing products.