Associate Professor Xiaoli Deng

Satellite altimetry has proven a valuable resource of information on river and lake levels where in situ data are sparse or non-existent. In this study several new methods for obtaining stable inland water levels from CryoSat-2 Synthetic Aperture Radar (SAR) altimetry are presented and evaluated. In addition, the possible benefits from combining physical and empirical retrackers are investigated.The retracking methods evaluated in this paper include the physical SAR Altimetry MOde Studies and Applications (SAMOSA3) model, a traditional subwaveform threshold retracker, the proposed Multiple Waveform Persistent Peak (MWaPP) retracker, and a method combining the physical and empirical retrackers. Using a physical SAR waveform retracker over inland water has not been attempted before but shows great promise in this study.The evaluation is performed for two medium-sized lakes (Lake Vänern in Sweden and Lake Okeechobee in Florida), and in the Amazon River in Brazil. Comparing with in situ data shows that using the SAMOSA3 retracker generally provides the lowest root-mean-squared-errors (RMSE), closely followed by the MWaPP retracker. For the empirical retrackers, the RMSE values obtained when comparing with in situ data in Lake Vänern and Lake Okeechobee are in the order of 2-5 cm for well-behaved waveforms. Combining the physical and empirical retrackers did not offer significantly improved mean track standard deviations or RMSEs. Based on these studies, it is suggested that future SAR derived water levels are obtained using the SAMOSA3 retracker whenever information about other physical properties apart from range is desired. Otherwise we suggest using the empirical MWaPP retracker described in this paper, which is both easy to implement, computationally efficient, and gives a height estimate for even the most contaminated waveforms.

Over inland waters the altimetry radar return is typically contaminated by both land and surrounding vegetation. Moreover, the radar returns corresponding to floodplain vegetation and exposed mud flats can have similar waveform shapes to those reflected from open water surfaces so that the derived water surface elevations are either inaccurate or difficult to extract. This letter presents a robust and automated method based on image analysis of satellite imagery for the selection of altimetry waveforms over inundated zones. The altimetry footprint is assessed as being inundated if the radiometric response of an image kernel within the footprint conforms to known remotely sensed responses for water. The method was applied in the lower middle Fly floodplain of the Fly River in the Western Province of Papua New Guinea (PNG). Altimetry waveforms were assessed for inundation extent and vegetation cover, with those that met threshold levels being flagged for further retracking and water surface elevation monitoring. The results show that the method accurately identified> 90% of inundated sites along altimeter ground tracks and correctly selected waveforms reflected from water surfaces.

A new approach for improving the accuracy of altimetry-derived sea level anomalies (SLAs) near the coast is presented. Estimation of SLAs is optimized using optimal waveform retracking through a fuzzy multiple retracking system and the most appropriate detiding method. With the retracking system, fuzzy-retracked SLAs become available within 5 km of the coast; meanwhile it becomes more important to use pointwise tide modelling rather than state-of-the-art global tidal models, as the latter leave residual ocean tide signals in retracked SLAs. These improvements are demonstrated for Jason-2 waveforms in the area of the Great Barrier Reef, Australia. Comparing the retrieved SLAs with in situ tide gauge data from Townsville and Bundaberg stations showed that the SLAs from this study generally outperform those from conventional methods, demonstrating that adequate waveform retracking and detiding are equally important in bringing altimetry SLAs closer to the coast. © 2014 Taylor & Francis.

It is broadly acknowledged that the precision of satellite-altimeter-measured instantaneous sea surface heights (SSH) is lower in coastal regions than in open oceans, due partly to contamination of the radar return from the coastal sea-surface state and from land topography. This study investigates the behavior of ERS-2 and POSEIDON altimeter waveform data in coastal regions and estimates a boundary around Australia's coasts in which the altimeter range may be poorly estimated by on-satellite tracking software. Over one million 20 Hz ERS-2 (March to April 1999) and POSEIDON (January 1998 to January 1999) radar altimeter waveform data were used over an area extending 350 km offshore Australia. The DS759.2 (5' resolution) ocean depth model and the GSHHS (0.2 km resolution) shoreline model were used together to define the coastal regions. Using the 50% threshold retracking points as the estimates of expected tracking gate, we determined that the sea surface height is contaminated out to maximum distance of between about 8 km and 22 km from the Australian shoreline for ERS-2, depending partly on coastal topography. Using the standard deviation of the mean waveforms as an indication of the general variability of the altimeter returns in the Australian coastal region shows obvious coastal contamination out to about 4 km for both altimeters, and less obvious contamination out to about 8 km for POSEIDON and 10 km for ERS-2. Therefore, ERS-2 and POSEIDON satellite altimeter data should be treated with some caution for distances less than about 22 km from the Australian coast and probably ignored altogether for distances less than 4 km.