SAR Radiometric Cross-Calibration Based on Multiple Pseudoinvariant Calibration Sites With Extensive Backscattering Coefficient Range

Synthetic aperture radar (SAR) radiometric cross-calibration achieves the calibration of uncalibrated satellites by employing calibrated satellites to illuminate the same ground targets. The stability of these ground targets is critical for effective cross-calibration. Such stable ground targets, named pseudoinvariant calibration sites (PICSs), have been extensively researched for optical sensors, but there has been comparatively less focus on SAR sensors. Furthermore, studies that synthesize multiple targets for comprehensive dynamic range calibration remain relatively unexplored. This article proposes an optimized method for selecting PICS for SAR radiometric cross-calibration to address these issues. This method incorporates spatial variation coefficients, spatial autocorrelation indicators, and edge extraction techniques to identify homogeneous targets. In addition, the backscattering coefficient's time-series root mean square error (RMSE) is leveraged to identify stable targets. Using time-series data from Sentinel-1, the study identifies 57 stable homogeneous targets as PICS, exhibiting a wide intensity distribution and a time-series RMSE less than 0.5 dB. Based on these PICS, this article further develops a wide dynamic range cross-calibration method based on multiple PICS sites. It employs weighted least squares regression for the different data quality of each site. The cross-calibration experiments demonstrate an average improvement in the accuracy of 0.82 dB compared to the cross-calibration method utilizing a single PICS site.