Abstract's details

EMD filtering applied to LRM 20 Hz sea level anomaly observations

Francesco Nencioli (CLS, France)


Marie-Isabelle Pujol (CLS, France); Nicolas Picot (CNES, France); Gerald Dibarboure (CNES, France); Yves Quilfen (IFREMER, France)

Event: 2022 Ocean Surface Topography Science Team Meeting

Session: Instrument Processing: Measurement and Retracking

Presentation type: Type Oral

Contribution: PDF file


Satellite radar altimetry is used to observe a wide range of spatial scales, ranging from basin scale to small mesoscale (i.e. less than 100 km). The analysis of the small mesoscale with low-resolution altimetry observations is usually based on along-track products, such as the Geophysical Data Record (GDR). Such product have a resolution as small as 300 m (20-Hz rate), although the resolution most commonly used is 6–7 km (1-Hz rate). Unfortunately, the observation of ocean scales smaller than 100 km with LRM products is degraded by the existence of a “hump artefact” visible on sea surface height (SSH) spectra of the 20 Hz product at scales between 3 to 100 km. This hump is due to inhomogeneities in backscatter strength within the LRM disc-shaped footprint, which induce retracking errors which are smoothed along the satellite track. Recent studies have evidenced that the impact of such spectral hump can be mitigated by more restrictive editing algorithm as well as by the use of high-frequency specific corrections, such as the high-frequency adjustment (HFA).

Here, we assess the effectiveness of the Empirical Mode Decomposition filter (EMD) in mitigating the impact of the hump. The EMD filter is a novel filtering method specifically designed for the analysis of non-stationary and non-linear signals. The filter is purely algorithmic (i.e. it lacks a sound mathematical theory) and breaks down the signal into a series of amplitude and frequency modulated zero-mean functions called Intrinsic Moulation Functions (IMF). The main difference with respect to traditional decomposition methods is that these functions are signal-dependent and are estimated via an iterative procedure. Once all IMFs are identified, the filtered signal is reconstructed by summing together only the significant portions (i.e. above a defined noise threshold) of each IMF.

The EMD method has been recently applied for de-noising 1 Hz significant wave height observations showing promising results. Here, we applied an analogous filter to 20 Hz sea level observations from Jason-3. Due to the higher resolution of our dataset, several different parameter configurations were tested (e.g. number of IMFs to include to reconstruct the filtered signal; type of iterative method;…). Our results confirmed that, in its optimal configurations, the filter is extremely efficient at removing the stochastic noise from the signal. However, the filter alone does not reduce the hump artefact observed in the 20 Hz spectra. Overall, the largest mitigation of the 3 to 100 km spectral hump was obtained using the EMD filtering in synergy with the HFA correction.

Oral presentation show times:

Room Start Date End Date
Sala Grande Tue, Nov 01 2022,11:30 Tue, Nov 01 2022,11:45
Francesco Nencioli