Multi-Frequency SAR Response to Crop Structure and Phenology in Western Canadian Fields: An Empirical Investigation of ICEYE X-Band and Sentinel-1 C-Band Imagery

Revised abstract:

Introduction: Synthetic aperture radar (SAR) enables weather-independent crop monitoring, but agricultural backscatter responds simultaneously to crop structure, soil moisture, weather, and acquisition geometry, so changes cannot be attributed to phenology alone. Because radar wavelength and polarization shape the response, X-band and C-band sensors observe the same field differently. This study investigates how ICEYE X-band and Sentinel-1 C-band dual-polarization imagery characterize crop structure and development in Western Canadian fields.

Methods: Field campaigns near Lethbridge, Alberta, during the 2025 growing season covered barley, canola, and spring wheat, with ground truth including field-verified phenology, biomass, soil moisture, and weather. ICEYE Stripmap acquisitions (VV, 2.5 m, 7- to 14-day revisit) were processed in SNAP through calibration to σ⁰, Lee Sigma speckle filtering, and Range-Doppler terrain correction. The same chain was applied to 26 Sentinel-1 GRD scenes (Interferometric Wide swath), using a 7 x 7 Lee Sigma filter and the SRTM DEM. Field-level VV and VH backscatter was summarized by crop in dB.

Results: ICEYE VV and Sentinel-1 VH showed opposite harvest responses: X-band VV increased by 1 to 4 dB as exposed soil and stubble raised surface scattering, whereas C-band VH decreased as canopy removal reduced volume scattering. Sentinel-1 VH separated canola from cereals, averaging about -18.3 dB against -21.5 dB for barley and -22.8 dB for spring wheat, while the cereals remained similar. Both sensors showed confound sensitivity: the X-band series included a shift near 6 dB tied to acquisition geometry rather than phenology, and cereal separability was reduced by backscatter overlap.

Conclusions: Because the two sensors differ in frequency and polarization, they can respond differently to the same event. ICEYE gave high-resolution structural detail and Sentinel-1 tracked multi-temporal development, yet both were sensitive to environmental and acquisition-geometry confounds. These findings motivate confound-aware analysis; ongoing work adds ground truth, seasons, and physical modeling.