A Novel Algorithm for Near Real Time Crop Progress Monitoring at Field Scales by Fusing Observations from Harmonized Landsat and Sentinel-2 and Geostationary Satellites

Crop phenology has been widely detected from multisource satellite observations, providing an unique opportunity to directly explore the anthropogenic impacts on the worldwide agroecosystem over long periods. Conversely, near real-time monitoring of crop progress by combining timely available and historical remote sensing observations is barely investigated. The main challenges are (1) the availability and frequency of cloud-free satellite observations, and (2) the flexibility and capacity of algorithms to capture and represent land-cover-induced background variations (climatology) within different growing seasons, such as crop rotations. Harmonized Landsat and Sentinel-2 (HLS) provides ground monitoring every 3~5 days at field-scale (30m), while Advanced Baseline Imagery (ABI) onboard a new generation of geostationary satellite observes croplands every 10-15 minutes at 0.5~1 km spatial-scale. The integration of HLS and ABI could offer high opportunities to improve the spatiotemporal capacity of currently available near real-time monitoring efforts from Visible Infrared Imaging Radiometer Suite (VIIRS) or Moderate Resolution Imaging Spectroradiometer (MODIS) observations. In this study, a novel algorithm was proposed for operationally producing weekly crop progress at a field-scale by fusing HLS and ABI observations (HLS-ABI) with a Spatiotemporal Shape-Matching Model (SSMM) throughout the growing season in near real time. Specifically, the two-band enhanced vegetation index (EVI2) time series from HLS-ABI in previous two years was used as the crop background information (reference temporal shape), representing the inter-annual variability of crop greenness and the spatially-distributed crop phenology variations induced by crop-rotations. The timely available HLS-ABI EVI2 time series, consisting of one-entire-year observations backwards from the current date, was built up by fusing HLS and ABI data with the SSMM algorithm. The timely available HLS-ABI EVI2 time series was then fused with an optimal reference temporal shape to predict the crop EVI2 after the current date. The timely available HLS-ABI EVI2 time series were further combined with the predicted future EVI2 to detect crop phenometrics of greenup onset, mid-greenup phase, maturity onset, senescence onset, mid-senescence phase, and dormancy onset. This near real time crop phenology monitoring was performed every week and the accuracy was evaluated by comparing with detections from standard HLS-ABI observations and near-surface PhenoCam observations in 2020. The near- and real time crop progress was further validated with the weekly released crop progress reports (CPRs) data from the National Agricultural Statistics Service (NASS) of the United States Department of Agriculture (USDA) in Iowa state in 2021