Crop height retrieval across different scenes with dual-pol backscattering and one-time in-situ data

Synthetic Aperture Radar (SAR) data, which can penetrate the canopy under all-weather conditions, holds great potential for crop height estimation. However, due to the rapid morphological changes in crops and their relatively low heights, current phase- or coherence-based methods require large spatial baselines and high temporal resolution data to estimate crop height. This paper proposes a novel SAR Height Index (SHI)-based crop height estimation algorithm that primarily relies on SAR dual-pol backscattering, while incorporating a logistic growth model to improve temporal consistency. The method avoids the need for interferometric spatial baselines and complex parameterization. First, SHI is introduced based on a Radial Basis Function (RBF) kernel with pixel-level adaptive σ calculation. Second, phenologically filtered SHI is regressed against observed crop heights to establish fixed, scene-specific empirical linear models for height estimation. Third, daily crop height maps are generated by integrating the estimated height and a logistic growth model that covers most of the growth cycle. Experiments were conducted on corn and soybean plantations in Spain using fully polarimetric ground-based SAR (GB-PolSAR) data, on a three-year winter wheat plantation in Germany, and a two-year cotton plantation in the USA, using Sentinel-1 (S1) GRD data. All results demonstrated a good correlation between SHI and the measured crop height. Daily height estimates for wheat and cotton derived from S1 showed close agreement with in situ measurements, with R 2 values ranging from 0.81 to 0.91 and RMSE values between 8.9 cm and 18.4 cm. Although the application of SHI across different climatic regions may require lightweight transfer calibration, it demonstrates greater generalizability than conventional backscattering or radar vegetation index. The SHI-based algorithm significantly enhances flexibility in data selection and provides a promising solution for region-scale and cost-effective crop height estimation. • Crop height estimation using only dual-pol SAR backscattering. • Reduces baseline limits for flexible, cost-effective height estimation. • Requires only one-time in-situ data per scene for model calibration. • Validated on four crops in Spain, Germany, and USA under different climates. • Tested with both ground-based PolSAR and Sentinel-1 data.