Liu et al. (2026) Crop height retrieval across different scenes with dual-pol backscattering and one-time in-situ data

Identification

• Journal: Remote Sensing of Environment
• Year: 2026
• Date: 2026-04-03
• Authors: Jinglong Liu, Jordi J. Mallorqui, Xavier Fàbregas, Antoni Broquetas, Albert Aguasca, Mireia Mas, Feng Zhao, Yunjia Wang
• DOI: 10.1016/j.rse.2026.115407

Research Groups

• CommSensLab, Department of Signal Theory and Communications (TSC), Universitat Politècnica de Catalunya (UPC), Barcelona, Spain
• School of Environment Science and Spatial Informatics, China University of Mining and Technology, Xuzhou, China
• Key Laboratory of Land Environment and Disaster Monitoring, MNR, China University of Mining and Technology, Xuzhou, China

Short Summary

This paper introduces a novel SAR Height Index (SHI)-based algorithm for crop height estimation that relies solely on dual-polarization SAR backscattering and a one-time in-situ calibration, demonstrating enhanced flexibility and cost-effectiveness by avoiding interferometric baseline constraints and complex parameterization. The method, validated across various crops and regions, integrates a logistic growth model to improve temporal consistency and fill data gaps, achieving high accuracy in daily crop height retrieval.

Objective

• To develop a novel SAR Height Index (SHI)-based crop height estimation algorithm that primarily relies on SAR dual-polarization backscattering, incorporates a logistic growth model for temporal consistency, and avoids the need for interferometric spatial baselines and complex parameterization.

Study Configuration

• Spatial Scale: Experimental fields (22 m × 60 m) in Barcelona, Spain (corn, soybean); Elementary Sampling Units (ESU) within 80 m × 80 m areas in Demmin, Germany (winter wheat); Two farms (Ashburn Cooperator Farm, Ty Ty Cooperator Farm) in Tifton, USA (cotton). Spatial resolutions: 1 m² for GB-PolSAR, 10 m for Sentinel-1 GRD.
• Temporal Scale: Continuous GB-PolSAR data at 10-minute intervals covering crop growth cycles (e.g., Corn: July-November 2020; Soybean: May-September 2023). Multi-year Sentinel-1 GRD data covering full growing seasons (Winter wheat: 2021-2023; Cotton: 2018-2019). Daily crop height maps generated by integrating a logistic growth model.

Methodology and Data

• Models used:
  • SAR Height Index (SHI) based on a Radial Basis Function (RBF) kernel.
  • Logistic growth model (sigmoid curve) for temporal regularization and gap filling.
  • Comparison with traditional approaches: individual polarimetric backscattering (𝜎◦𝑣𝑣, 𝜎◦𝑣ℎ), Radar Vegetation Index (RVI).
  • Conceptual comparison with Interferometric SAR (InSAR) and Polarimetric InSAR (PolInSAR) methods (e.g., RVoG, OVoG).
• Data sources:
  • Ground-based Fully Polarimetric SAR (GB-PolSAR) data (C-band, HH, HV, VV) from Barcelona, Spain.
  • Sentinel-1 (S1) GRD data (C-band, VV, VH) from Google Earth Engine (GEE) for Demmin, Germany, and Tifton, USA.
  • In-situ crop height measurements for calibration (one-time per scene) and validation (multi-temporal).
  • Meteorological records for environmental context.

Main Results

• The proposed SHI demonstrated a good correlation with measured crop height for all four crops (corn, soybean, winter wheat, cotton), with R² values exceeding 0.77.
• SHI derived from VV and HV polarizations consistently outperformed HH and HV combinations, attributed to VV's sensitivity to stable vertical crop structures.
• Integration of the logistic growth model (LGM) effectively reconstructed daily crop height estimates, compensating for data gaps and improving temporal continuity.
• LGM-enhanced results showed improved accuracy:
  • Winter wheat (S1 GRD): R² values ranged from 0.81 to 0.91, with RMSE between 8.96 cm and 12.13 cm.
  • Cotton (S1 GRD): R² consistently exceeded 0.73, with RMSE values ranging from 12.94 cm to 18.37 cm.
• The bounded adaptive strategy for sigma (RBF kernel width) estimation achieved the highest accuracy among tested strategies.
• The SHI framework proved robust to input uncertainties, with maximum R² deviations below 0.14 and RMSE differences below 3.5 cm in Monte Carlo simulations.
• SHI-based models exhibited superior generalizability across years and sites (within regional scales) compared to models using individual backscattering coefficients or RVI.

Contributions

• Introduction of a novel SAR Height Index (SHI) based on an RBF kernel, utilizing dual-polarization backscattering to estimate crop height without requiring interferometric spatial baselines.
• Development of a flexible and cost-effective crop height estimation algorithm that minimizes data requirements, needing only one-time in-situ data per scene for calibration.
• Integration of a logistic growth model to enhance temporal consistency and fill data gaps, enabling the generation of daily, biologically plausible crop height estimates.
• Demonstration of improved generalizability of SHI-based models across diverse crops, years, and agro-climatic regions compared to conventional backscattering or radar vegetation index approaches.
• Significant reduction in data burden and computational cost for crop height retrieval, offering a promising solution for region-scale monitoring.

Funding

• Project PID2024-161188OB-C21 (Spanish Ministry of Science, Innovation and Universities MICIU/AEI/10.13039/501100011033 and European Regional Development Fund FEDER, UE)
• Doctoral grant PRE2021-097981
• National Natural Science Foundation of China (Grant No. 42474018)
• National Key R&D Program of China (Grant No. 2022YFE0102600)
• Construction Program of Space-Air-Ground-Well Cooperative Awareness Spatial Information Project (B20046)

Citation

@article{Liu2026Crop,
  author = {Liu, Jinglong and Mallorqui, Jordi J. and Fàbregas, Xavier and Broquetas, Antoni and Aguasca, Albert and Mas, Mireia and Zhao, Feng and Wang, Yunjia},
  title = {Crop height retrieval across different scenes with dual-pol backscattering and one-time in-situ data},
  journal = {Remote Sensing of Environment},
  year = {2026},
  doi = {10.1016/j.rse.2026.115407},
  url = {https://doi.org/10.1016/j.rse.2026.115407}
}