Zhou et al. (2025) 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

Identification

• Journal: International Journal of Applied Earth Observation and Geoinformation
• Year: 2025
• Date: 2025-12-03
• Authors: Qu Zhou, Kaiyu Guan, Sheng Wang, Xiaocui Wu, Samuel Stroebel, James D. Hipple
• DOI: 10.1016/j.jag.2025.105004

Research Groups

• Agroecosystem Sustainability Center, Institute for Sustainability, Energy, and Environment, University of Illinois Urbana-Champaign, Urbana, IL, USA
• Department of Natural Resources and Environmental Sciences, College of Agricultural, Consumer and Environmental Sciences, University of Illinois Urbana-Champaign, Urbana, IL, USA
• National Center for Supercomputing Applications, University of Illinois Urbana-Champaign, Urbana, IL, USA
• Department of Computer Science, University of Illinois Urbana-Champaign, Urbana, IL, USA
• Indiana State Department of Agriculture, Indianapolis, IN, USA
• United States Department of Agriculture, Risk Management Agency, Washington, DC, USA

Short Summary

This study evaluated the performance of multi-sensor satellite vegetation time series (HLS, MODIS, PlanetScope) for phenology-based field-level cover cropping detection in Indiana. It found that Harmonized Landsat-8 and Sentinel-2 (HLS) outperformed MODIS and MODIS-calibrated PlanetScope, with original PlanetScope showing the highest accuracy, and identified key factors influencing detection accuracy.

Objective

• To evaluate the performance of Normalized Difference Vegetation Index (NDVI) time series from satellite sensors with varied spatial, temporal, and radiometric characteristics in phenology-based field-level cover cropping detection.
• To leverage multi-scale ground truth cover cropping data and quantify their discrepancies for phenology-based field-level cover cropping detection.
• To analyze the impacts of cover cropping-related factors (field sizes, regional adoption rates, and cover crop species) on phenology-based field-level cover cropping detection.

Study Configuration

• Spatial Scale: Field-level detection across Indiana, USA, with satellite data resolutions of 3 meters (PlanetScope), 30 meters (HLS), and 250 meters (MODIS).
• Temporal Scale: The primary study period was 2017, with NDVI consistency evaluated for 2017–2018. Satellite data temporal resolutions included approximately 3-day (HLS) and daily (MODIS, PlanetScope).

Methodology and Data

• Models used:
  • Phenology-based cover cropping detection framework (adapted from Zhou et al., 2022).
  • Revised Savitzky-Golay filter-based algorithm for NDVI time series gap-filling and noise reduction.
  • Logistic regression model to quantify discrepancies between NASS and ISDA county-level adoption rates.
  • One-way ANOVA for assessing performance differences.
  • Bootstrap sampling with 500 replicates for uncertainty estimation.
• Data sources:
  • Satellite:
    • Harmonized Landsat-8 and Sentinel-2 (HLS v1.4) NDVI time series (30 m, ~3-day).
    • MODIS (MOD09GQ and MYD09GQ) NDVI time series (250 m, daily).
    • PlanetScope NDVI time series (3 m, near-daily), including MODIS-calibrated PlanetScope (CPS).
    • MODIS NBAR (MCD43A4) for PlanetScope radiometric cross-calibration (500 m, daily).
  • Observation/Ground Truth:
    • USDA National Agricultural Statistics Service (NASS) Census of Agriculture (county-level cover cropping acreage).
    • Indiana State Department of Agriculture (ISDA) roadside transect surveys (field-level cover cropping presence, species, residue, prior cash crops).
  • Auxiliary:
    • NASS Cropland Data Layer (CDL) (30 m) for crop type information (corn, soybean).
    • Internal Field Boundary Layer (based on Common Land Unit).

Main Results

• HLS outperformed MODIS and MODIS-calibrated PlanetScope (CPS) in phenology-based field-level cover cropping detection, achieving an average accuracy of 76.2 % ± 3.0 % across Indiana in 2017.
• Original PlanetScope achieved the highest average detection accuracy of 81.8 % in five selected counties, compared to 66.0 % for CPS, indicating that MODIS-based radiometric calibration reduced PlanetScope’s sensitivity to small NDVI changes.
• Significant discrepancies between ISDA and NASS county-level cover cropping adoption rates were found in 9 % of counties. These discrepancies were negatively correlated with detection accuracies (r = -0.54, p < 0.005) in counties where ISDA-reported adoption was substantially higher than NASS.
• Detection accuracy was higher in larger cover cropping fields; detected fields had a median size of approximately 9.36 hectares (130 HLS pixels), while non-detected fields had a median size of approximately 3.6 hectares (50 HLS pixels).
• Detection accuracy was positively correlated with regional cover cropping adoption rates (r = 0.42, p < 0.001).
• Detection accuracy varied by cover crop species: wheat (88.95 %), winter grains (77.73 %), ryegrass (75.52 %), barley (75.35 %), and cereal rye (68.59 %).

Contributions

• Provided a comprehensive multi-sensor assessment of HLS, MODIS, and PlanetScope NDVI time series for phenology-based field-level cover cropping detection.
• Demonstrated the superior performance of HLS and original PlanetScope for detecting cover crops, highlighting the importance of balanced spatial and temporal resolution and the potential drawbacks of coarse-resolution calibration for high-resolution data.
• Leveraged and quantified discrepancies between multi-scale ground truth data (NASS county-level and ISDA field-level) and analyzed their impact on detection accuracy.
• Identified and analyzed the influence of cover cropping field sizes, regional adoption rates, and cover crop species on detection accuracy, offering valuable insights for improving cover cropping monitoring and conservation programs.

Funding

• USDA National Institute of Food and Agriculture (NIFA) Foundational Program awards
• US Department of Energy’s Advanced Research Projects Agency-Energy (ARPA-E) SMARTFARM (MBC Lab and SYMFONI) projects
• NSF CAREER Award by Environmental Sustainability Program
• C3.ai Digital Transformation Institute
• NASA Early Career Investigator Program in Earth Science (80NSSC24K1057)
• NASA FINESST award
• ESA NoR (Network of Resources) Sponsorship

Citation

@article{Zhou2025Multisensor,
  author = {Zhou, Qu and Guan, Kaiyu and Wang, Sheng and Wu, Xiaocui and Stroebel, Samuel and Hipple, James D.},
  title = {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},
  journal = {International Journal of Applied Earth Observation and Geoinformation},
  year = {2025},
  doi = {10.1016/j.jag.2025.105004},
  url = {https://doi.org/10.1016/j.jag.2025.105004}
}