Abstract's details

Performances of a stratified and neural network combined approach for the wet tropospheric correction retrieval

Marie-Laure Frery (CLS, France)


Bruno Picard (CLS, France); Mathilde Siméon (CLS, France); Estelle Obligis (CLS, France); Laurence Eymard (LOCEAN/IPSL, France); Sophie Le Gac (CNES, France)

Event: 2016 Ocean Surface Topography Science Team Meeting

Session: Instrument Processing: Corrections

Presentation type: Type Poster

Contribution: PDF file


The wet tropospheric correction (WTC) is a major source of uncertainty in altimetry budget error, due to its large spatial and temporal variability: this is why the main altimetry missions include a microwave radiometer (MR) The commonly agreed requirement on WTC for current missions is to retrieve WTC with an error better than 1cm rms.

JPL for NASA/CNES (Jason-2) missions on one hand and CLS/IPSL for CNES (AltiKa) and ESA (Sentinel-3) missions on the other hand based their retrievals on similar approaches with still identified differences.

Both are based on an empirical relation relating the top of atmosphere brightness temperatures (TB) to the WTC.

With CLS approach, this relation is established using ECMWF analysis for the reference WTC and the inputs of the radiative transfer model used to compute simulated TB. A single neural network (NN) is used to invert the relation at a global scale.

With JPL approach, radiosondes are used to compute the reference WTC. A Log-linear parametric model is used to invert the relation in a stratified scheme: different coefficients of the log-linear model are established for different ranges of wind speed and WTC.

In the frame of CNES project PEACHI-J3, we defined a combined approach using the ECMWF analysis and the neural network in a stratified scheme.

Stratified-NN WTC is computed for Jason-3 radiometer and compared to the WTC in operational products. Performances are discussed and conclusions are drawn for future operational retrievals.

Poster show times:

Room Start Date End Date
Grande Halle Thu, Nov 03 2016,11:00 Thu, Nov 03 2016,18:00
Marie-Laure Frery