The airborne multi-channel radar depth sounder systems (MCRDS) are developed at the Center for Remote Sensing of Ice Sheets (CReSIS) to map the ice-sheet bed, deep internal layers, and fast-flowing outlet glaciers with their high sensitivity for weak echoes and the beamforming ability of the receive antenna array for clutter reduction. The work of this dissertation is part of the efforts CReSIS put in to obtain the best results in processing the data MCRDS radars have collected in Greenland. This dissertation includes the waveforms design, the development and implementation of SAR processing, and clutter reduction algorithms for MCRDS radars.;Big ice attenuation is the greatest challenge in sounding fast outlet glaciers using airborne radars. Besides the elaborate hardware design, MCRDS radars maximize the sensitivity to overcome the signal loss by system calibration, channel mismatch compensation, RFI suppression and SAR (synthetic aperture radar) processing with aircraft motion compensations in data processing. In this dissertation, the SNR gains from the calibration of reference functions for pulse compression and the compensation of constant channel mismatches are verified with echoes from the ice bed. Some deep ice layers of the Greenland ice sheet that are masked by RFI (radio frequency interference) in MCRDS data are revealed by applying adaptive array processing. A SAR algorithm based on wavefront reconstruction theory is developed and implemented in frequency and wave-number domains with narrow beamwidth motion compensation. The SNR gains by SAR processing with motion compensation are carefully verified by using simulation data and sea-ice data. With the verified SAR processing algorithms, very weak echoes from the deepest parts of Jakobshavn channel are detected for the first time using large synthetic aperture length in radar soundings and the depths match with seismic measurements.;While SAR processing effectively reduces along-track clutter, across-track clutter is another challenge encountered in sounding fast-flowing outlet glaciers. MCRDS radars facilitate rejection of across-track surface clutter with small arrays of four, five or six elements. In this dissertation, three clutter-reduction algorithms are either developed or implemented: (1) the data-dependent FMV algorithm, (2) the data-independent null-steering algorithm, and (3) the clutter-power estimation algorithm. The first two algorithms reduce clutter signals by 34.30 dB and 28.57 dB respectively when applied to sea-ice data. But neither is very effective when applied to ice data with distributed clutter. The third algorithm developed is a beam-spaced method. It is more robust to channel the mismatches and clutter angle estimate errors that are the limiting factors of the first two methods. There are two stages of the beam-spaced method. The first stage is to form a main beam and a clutter beam. The main beam is formed by choosing weights to enhance the nadir signals and with clutter signals partly reduced. The clutter beam is formed by choosing weights to put a null at nadir and to have maximum gains in the direction of clutter. The second stage is to subtract the weighted clutter beam from the main beam to properly compensate the gain difference between the two beams based on power profiles estimation. Two clutter scenarios are used to illustrate the effectiveness of the beam-spaced algorithm. In the first scenario, the aircraft's altitude is high and the ice bed masked by clutter is deep, while in the second case the aircraft's altitude is low and the ice bed masked by clutter is shallow. In both scenarios the beam-spaced algorithm reduces clutter further beyond the reduction by Hanning weighting. The further clutter reduction is around 10.3 dB in the first case and 9.6 dB in the second one. This dissertation also presents the results of applying the beam-spaced method in two cases over Jakobshavn channel. In the first case, the across-track ice clutter is cleared but the channel ice bed is still invisible because of the huge ice attenuation in the channel. In the second case that needs to be further studied, the method fails to reduce the clutter-like signals near the channel's calving front. (Abstract shortened by UMI.)
展开▼
