A 4D look–bin–space–time (LBiST) approach for waveform decontamination and range determination in SAR altimetry

Accurate range determination in satellite nadir altimetry over inland waters remains limited by the way Level-2 waveforms are generated. The multilooking stage in SAR processing mixes echoes from water, surrounding land, and secondary water returns across the altimeter footprint. The generated waveform is often distorted or multi-peak, which hinders range estimation, particularly over narrow rivers, irregular shorelines, or partially filled reservoirs. In such cases, even advanced retrackers struggle to identify the true retracking line. This study introduces a novel approach tailored for challenging inland environments by leveraging the spatial-temporal information embedded in Sentinel-3 SRAL Level-1BS (L1BS) SAR data products. Instead of conventional waveform analysis for retracking, we directly utilize the stack of high-resolution complex echoes, organized in look-bin maps, which are available after slant range correction and before multilooking in the SAR processing scheme. A Bayesian inference framework is employed, where priors are constructed across cycles (time domain) to capture meaningful temporal patterns, such as persistent water-surface signatures. Likelihoods are extracted from stacked look-bin maps within each cycle (space domain), enabling suppression of non-water and secondary water reflections. The resulting posterior probability highlights regions in the look-bin maps most likely associated with echoes from the nadir water surface (nearest point on the surface to satellite): nadir-water echo zone in look-bin maps . Using this formulation, our approach effectively exploits the four dimensions of Look, Bin, Space, and Time (LBiST) of the return signal. This enables waveform decontamination, resulting in improved range estimates that account for angle-dependent reflectance and spatiotemporal continuity. We evaluate the method over challenging inland-water environments, including narrow rivers, fragmented reservoirs, and complex shorelines where applying conventional retrackers on standard Level-1B waveforms often underperforms. Without applying post-retrieval outlier rejection, the LBiST approach consistently improved water-level estimation relative to the operational retracker OCOG, with mean correlation/RMSE improving from 0.49/1.56 m to 0.79/0.49 m for all measurements across cycles, and from 0.67/0.93 m to 0.86/0.44 m for cycle medians. Overall, LBiST reduced outliers substantially and generally outperformed the OCOG, ICE, Threshold, SAMOSA, and SAMOSA + retrackers in contaminated inland waveforms.