A multi-parameter optimized sub-waveform retracker for monitoring river water levels using SAR altimetry

Abstract

Synthetic Aperture Radar (SAR) altimetry has been widely used for monitoring river water levels, especially over large and medium-sized rivers. However, challenges still remain in obtaining continuous and high-precision water levels over small rivers due to the altimeter's sparse along-track sampling, distorted waveforms, and river slopes. This study presents a new multi-parameter optimized sub-waveform (MulPOS) retracker, which retracks the waveforms across all cycles through a quantitatively considered integration of the spatial consistency and temporal continuity of water levels, river slopes, and the strong reflectivity of the river surface. Firstly, along-track sampling is increased by searching for off-nadir observations within half the sampling resolution from nadir water bodies to retrieve continuous river water levels. Secondly, waveform preprocessing, including interpolation and filtering is used to determine more accurate retracking points, and then all possible sub-waveform sets that correspond to river reflections are formed. The most likely sub-waveform sets are determined by their four-parameter weighting function, which considers spatial consistency, temporal continuity of water level variations, and the high reflectivity of the river water surface. Finally, slope corrections are computed using the robust Helmert variance component estimation method by combining the differences between water levels in adjacent cycles and along the track. The MulPOS has been applied to 290 virtual stations formed by Sentinel-3A/3B and Sentinel-6 MF over rivers in the United States (52 % of which are narrower than 100 m). For comparison purposes, six other retrackers have been used, including OCOG, ICE1, threshold, NPPTR, SAMOSA+, and MWaPP+. The results have been validated against the in-situ measurements from the United States Geological Survey, indicating that the water levels derived by MulPOS are superior to other retrackers with a median RMSE of 17.9 cm, a median relative RMSE of 7.2 %, a median correlation coefficient of 0.96, and an abnormal water level occurrence rate of 0.60 %, whereas the corresponding metrics for other retrackers are >24.2 cm, >9.8 %, <0.94, and > 2.36 %. Moreover, MulPOS achieves steady and high-precision water levels across most small rivers under varying river widths (e.g., RMSE for MulPOS is 20.9 cm vs. >29.5 cm for other retrackers over rivers narrower than 50 m), varying angles between satellite ground tracks and rivers, and complex river morphologies. MulPOS is expected to generate a dataset with continuous, high-precision water level data for more small and medium-sized rivers, and this will expand the application of altimetry to inland water monitoring.