High-Precision Assessment and Calibration of Polarimetric RADARSAT-2 SAR Using Transponder Measurements

Abstract

Independent assessment and calibration of polarimetric RADARSAT-2 (RS2) synthetic aperture radar (SAR) are conducted using “uncalibrated” data collected at various incidence angles (from 20 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">°</sup> to 40 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">°</sup> ). Analysis of the response of a transponder deployed at HV and VH configurations permits high-precision measurement of RS2 antenna crosstalks. It is shown that the RS2 antenna is highly isolated (better than -32 dB) with crosstalks stable with incidence angle. A new calibration method based on transponder measurements is introduced. It is shown that the transponder calibration method removes almost completely the low antenna crosstalk with a residual crosstalk lower than -43 dB. Only one transmitter-receiver distortion matrix measured at a given incidence angle is required for accurate calibration of the 20 polarimetric modes of the RS2 SAR from 20 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">°</sup> to 40 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">°</sup> . Uncalibrated RS2 single-look complex (SLC) data with real and imaginary parts provided in 32-bit floating point are required for an effective application of the transponder calibration method. RS2 “calibrated” data are also considered for the assessment of the actual RS2 polarimetric calibration. It is shown that RS2 calibration meets comfortably the CEOS Cal-Val requirements with a residual crosstalk lower than -32 dB. The RS2 calibration accuracy does depend on the mode (i.e., incidence angle) with residual crosstalk that varies between - 32 and -43 dB. To assess the impact of the residual crosstalk on polarimetric applications, the RS2 data are recalibrated using transponder measurements. Each data set was processed four times with a different lookup table (LUT) to reconstruct the 32-bit floating-point data prior to the application of the transponder calibration method. It is shown that the recalibration may not be required for natural targets of relatively high HV backscattering (higher than - 26 dB), such as forests. Data recalibration improves significantly the accuracy of targets of low HV backscattering measurements, whereas like polarization (HH and VV) does not seem to be affected by the residual antenna crosstalk (-32 dB). Urban targets that manifest significant helicity scattering may have their like polarization affected and require data recalibration. However, RS2 data recalibration requires the deployment of a transponder (at HV and VH polarization) for each mode (i.e., incidence angle). Data symmetrization is introduced as a more convenient way to improve polarimetric RS2 data quality without the need to deploy transponders. It is shown that the symmetrization of the modes with the highest residual crosstalk significantly improves the calibration accuracy with a residual crosstalk lower than -37 dB. The latter is negligible as the RS2 noise floor of β <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">°</sup> lies between -30 and -34 dB.