The impacts of cover crop performance on satellite-based detectability of cover crops in the Mid-Atlantic

Authors: XU, YIDE - University Of Illinois Urbana-Champaign; ZHOU, QU - University Of Illinois Urbana-Champaign; GUAN, KAIYU - University Of Illinois Urbana-Champaign; WANG, SHENG - University Of Illinois Urbana-Champaign; HIVELY, W. DEAN - Us Geological Survey; Jennewein, Jyoti; THIEME, ALISON - University Of Maryland; Mirsky, Steven; CHEN, ZHANGLIANG - University Of Illinois Urbana-Champaign

Submitted to: International Journal of Applied Earth Observation and Geoinformation
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/1/2026
Publication Date: 7/15/2026
Citation: Xu, Y., Zhou, Q., Guan, K., Wang, S., Hively, W., Jennewein, J.S., Thieme, A., Mirsky, S.B., Chen, Z. 2026. The impacts of cover crop performance on satellite-based detectability of cover crops in the Mid-Atlantic. International Journal of Applied Earth Observation and Geoinformation. 152. Article e105453. https://doi.org/10.1016/j.jag.2026.105453.
DOI: https://doi.org/10.1016/j.jag.2026.105453

Interpretive Summary: Cover crops are increasingly used to improve soil health and reduce nutrient losses, but quantifying their biomass is essential to fully understand their environmental benefits. This study used unique field-level datasets from Maryland and time series satellite imagery (2017–2021) to evaluate how cover crop biomass, planting, and termination dates affect their detectability with remote sensing. Time-integrated greenness metrics developed from satellite imagery showed that time series normalized difference vegetation index (NDVI) features were most effective index for estimating biomass. Results revealed that cover crops with higher biomass were more easily detected, with detection accuracy reaching 96.1% for biomass above 500 kg ha-1 compared to 62.7% overall. In other words, cover crops with at least 500 kg ha-1a were more reliably detectable compared to those with less accumulated biomass. Management practices strongly influenced outcomes, as earlier planting and later termination increased biomass at a rate of 4.14 kg ha-1 per day, thereby enhancing detectability. Researchers and administrators in federal, state, and university programs could use these findings to improve cover crop detection algorithms, enhancing cover crop monitoring programs and enabling policymakers to track progress on cover crop adoption and soil conservation efforts that support farmers and agricultural systems.

Technical Abstract: Cover crop adoption in the U.S. has greatly increased over the past several decades. Quantifying their biomass is essential for evaluating associated environmental benefits. While remote sensing (RS) based approaches for detecting cover crop presence have been developed, there has been limited research on how varying cover crops biomass levels and management practices influence detectability. Using unique datasets of field-level cover crop availability and biomass in Maryland, U.S. we investigated how RS-based detectability changes for cover crops with varied aboveground biomass and planting and termination dates. Specifically, we developed a time-integrated satellite-based greenness feature from Harmonized Landsat-8 and Sentinel-2 (HLS) time series from 2017 to 2021 to estimate biomass of cover crops and evaluate their detectability using a phenology-based cover crop detection framework. The impacts of cover crop planting and termination dates on cover crop biomass and detectability were also analyzed. Our results indicate that Normalized Difference Vegetation Index (NDVI) outperforms other indices such as Near-Infrared Reflectance of Vegetation (NIRv), Enhanced Vegetation Index (EVI), and Green Chlorophyll Vegetation Index (GCVI) in estimating cover crop biomass. Our time-integrated model achieved better performance than the conventional single-date “snapshot” linear model by improving R2 from 0.53 to 0.66 and reducing Root Mean Squared Error (RMSE) from 922 kg/ha to 747 kg ha-1. Moreover, while the snapshot model was sensitive to cover crop species, the time-integrated model showed strong robustness across different species. Detectability increased with cover crop biomass, as detected cover crops (963.3 ± 719.5 kg ha-1) had higher biomass than non-detected cover crops (297.2 ± 209.0 kg ha-1). The detection accuracy for cover crops with biomass larger than 500 kg ha-1 was 96.1%, significantly higher than the overall detection accuracy of 62.7% when the full range of biomass were included. Cover crops planted earlier and terminated later resulted in higher biomass and RS-based detectability, likely because biomass increased with longer growth duration at a rate of 4.14 kg ha-1 per day (p < 0.01). This study underscored the importance of management practices on cover crop biomass and their detectability via satellite time series, offering valuable insights for better managing cover crops and monitoring their impacts on agroecosystems.