Multi-sensor assessment of phenology-based field-level cover cropping detection using satellite vegetation time series from Harmonized Landsat-8 and Sentinel-2, MODIS, and PlanetScope

• Multi-sensor assessment of phenology-based field-level cover cropping detection. • HLS surpasses MODIS and MODIS-calibrated PlanetScope with an accuracy of 76%. • MODIS-based radiometric calibration decreases PlanetScope’s performance. • Multi-scale cover cropping data are leveraged with quantified discrepancies. • Field size, regional adoption, and cover crop species impact detection accuracy. Remote sensing-detected phenological differences among cover crops, cash crops, and soil backgrounds are crucial for detecting cover cropping practices. However, variations in detection performance with satellite data of different spatial, temporal, and radiometric characteristics and the potential factors affecting the performance remain unclear. Building on the previously developed framework, we aimed to evaluate the performance of multiple satellite sensors, leverage multi-scale ground truth data, and identify potential factors for improving field-level cover cropping detection. Specifically, we (1) compared widely-used NDVI time series from Harmonized Landsat-8 and Sentinel-2 (HLS, ∼3-day and 30 m), MODIS (daily and 250 m), and PlanetScope (near-daily and 3 m, requiring radiometric calibration), (2) quantified discrepancies between county-level data from National Agricultural Statistics Service (NASS) and field-level data from Indiana State Department of Agriculture (ISDA), and (3) analyzed the impacts of cover cropping field sizes, regional adoption rates, and cover crop species on cover cropping detection. We found that (1) HLS outperformed MODIS and MODIS-calibrated PlanetScope (CPS) with an average accuracy of 76.2 % ± 3.0 % across Indiana in 2017. (2) Significant discrepancies between ISDA and NASS data were found in 9 % of counties and were negatively correlated with detection accuracies in counties where ISDA-reported adoption was substantially higher than that reported by NASS (r = -0.54, p < 0.005). (3) Detection accuracy was higher in larger cover cropping fields, positively correlated with regional adoption rates (r = 0.42 and p < 0.001), with the highest accuracy for wheat (88.95 %), followed by winter grains (77.73 %), ryegrass (75.52 %), barley (75.35 %), and cereal rye (68.59 %). Our study offers valuable insights into selecting satellite sensors, reconciling multi-scale ground truth data, and identifying potential factors for improving phenology-based field-level cover cropping detection.